Developers

Read the Antseed marketplace by API or MCP.

curl https://antseedstats.com/api/v1/network/summary
54 endpoints54 MCP toolsNo key60 req/minCORS open
Base URL
antseedstats.com/api/v1
Auth
None
Rate limit
60 req/min per IP
Format
JSON, CSV on series
CORS
Open to any origin
OpenAPI
Reading a responseThe envelope, the units and who says what
{
  "data": { ... },
  "meta": {
    "version": "v1",
    "endpoint": "sellers.daily",
    "as_of_block": 52301234,
    "as_of_ts": 1791400000,
    "indexer_lag_blocks": 3,
    "generated_at": 1791400010,
    "max_age_s": 300,
    "sources": { "days": "chain" }
  }
}
  • as_of_block
    The Base block our indexer had fully processed when the response was built, with its timestamp and how many blocks behind the chain head that was. A number without a block cannot be reconciled with the chain; this one can.
  • version
    Fields never change meaning inside v1; new fields are added, old ones stay. Figures that depend on a definition of ours, such as organic or the trust score, say so in meta.sources, and a change of definition would be a new field or a v2.
  • money
    Money comes twice: an integer in micro-USDC (6 decimals, exact, the on-chain unit) and a formatted string with 2 decimals. Tokens and requests are integers. Prices are USD per 1 million tokens, as advertised.
  • time
    Timestamps are UTC seconds (fields ending in _ts). Days are UTC calendar days as YYYY-MM-DD. A 24h, 7d or 30d window is rolling and ends now; a day is a full UTC day, and today is partial.
  • sources
    chain: read from Base events or contract state. announce: what a seller signs and publishes about itself (names, prices), taken as is. antseedstats: a figure that depends on a definition of ours, such as the organic view, the trust score or a projection.
  • view
    Lists of sellers and prices default to all: nothing hidden. Rankings default to organic, like the site: sellers in Antseed's on-chain wash registry and in clusters we confirmed are left out. Every row still carries wash_registry (the chain's verdict) and organic (ours) so you can choose.
MCP for AI clients
MCPConnect an AI clientClaude, Cursor, VS Code, or any MCP client

The server runs at https://antseedstats.com/mcp, no key. It exposes the 54 endpoints below as tools and answers with the same data, each result stamped with the page it comes from and the Base block it is as of.

In a terminal
claude mcp add --transport http antseedstats https://antseedstats.com/mcp

Open source: @antseedstats/mcp-server on npm, source on GitHub. Same 60 req/min per IP as the API.

MCPTools (54)One per endpoint, with a question to try
  • antseedstats_network_health/network/health
    “Which sellers report failing services right now, and why?”
  • antseedstats_network_summary/network/summary
    “How much did the Antseed network settle today and this week?”
  • antseedstats_network_daily/network/daily
    “Chart Antseed's daily settled volume for the last 30 days”
  • antseedstats_leaders/leaders
    “Who is leading Antseed in the last 24 hours, by sales and by tokens?”
  • antseedstats_distribution/distribution
    “How concentrated is Antseed's volume among buyers and sellers?”
  • antseedstats_sellers/sellers
    “List the top 10 Antseed sellers by revenue this week, wash traders left out”
  • antseedstats_sellers_detail/sellers/{address}
    “Profile seller 0x1f22...18c9: revenue, trust score, verification, wash flag”
  • antseedstats_sellers_daily/sellers/{address}/daily
    “How much did seller 0x1f22...18c9 settle per day over the last week?”
  • antseedstats_sellers_checks/sellers/{address}/checks
    “Which of this Antseed seller's models failed the model audits, and by how much?”
  • antseedstats_sellers_models/sellers/{address}/models
    “Which models does seller 0x1f22...18c9 advertise, at what price, and which sold?”
  • antseedstats_registry/registry
    “Which sellers are in Antseed's on-chain wash-trading registry?”
  • antseedstats_buyers_top/buyers/top
    “How many buyers paid on Antseed this week, and who spent the most?”
  • antseedstats_buyers_days/buyers/daily
    “How many buyers were active on Antseed each day last month, and how many were new?”
  • antseedstats_buyers_history/buyers/history
    “How many buyers has Antseed had since the start, and how many still buy?”
  • antseedstats_buyers_free_to_paid/buyers/free-to-paid
    “How many Antseed free-tier users go on to pay, and which sellers' free tiers bring them?”
  • antseedstats_buyers/buyers
    “Who are the top 10 buyers on Antseed by lifetime spend?”
  • antseedstats_buyers_escrow/buyers/escrow
    “How much USDC have Antseed buyers deposited into escrow, and how much is still there?”
  • antseedstats_buyers_detail/buyers/{address}
    “Profile buyer 0x3f1d...f0f5: spend, sellers used, models bought”
  • antseedstats_buyers_daily/buyers/{address}/daily
    “How much did buyer 0x3f1d...f0f5 spend per day over the last week?”
  • antseedstats_buyers_rewards/buyers/{address}/rewards
    “How much $ANTS has buyer 0xa193...25ac earned, claimed and staked, and where?”
  • antseedstats_models/models
    “Which models are most used on Antseed, and what do they cost?”
  • antseedstats_models_checks/models/{id}/checks
    “Which Antseed sellers of claude-fable-5.1 really serve the model?”
  • antseedstats_models_video/models/video
    “How much do buyers spend on video on Antseed, and which video models sell most?”
  • antseedstats_models_images/models/images
    “How much do buyers spend on image generation on Antseed, and which image models sell most?”
  • antseedstats_models_speech/models/speech
    “How much do buyers spend on text-to-speech on Antseed, and which speech models sell most?”
  • antseedstats_models_detail/models/{id}
    “Where can I buy kimi-k3 on Antseed, cheapest trusted seller first?”
  • antseedstats_prices/prices
    “Compare every seller's price for kimi-k3 against the OpenRouter list price”
  • antseedstats_labs/labs
    “Which AI labs have models for sale on Antseed, and how many each?”
  • antseedstats_labs_detail/labs/{org}
    “Which Moonshot models are on Antseed and what is the cheapest price for each?”
  • antseedstats_staking_pools/staking/pools
    “Which Antseed provider pools hold the most stake and power, and what do they earn?”
  • antseedstats_staking_free/staking/free
    “Which Antseed provider gives me the most free AI for 20,000 ANTS staked for a year?”
  • antseedstats_staking_positions/staking/positions
    “List the open staking positions in pool 7 with their lock terms”
  • antseedstats_rewards_current/rewards/current
    “Who is earning the most $ANTS this epoch on Antseed, and how much is projected?”
  • antseedstats_rewards_pools/rewards/pools
    “How many active Antseed sellers hold a staking pool, and how much is locked and staked?”
  • antseedstats_referrals/referrals
    “Who are the top Antseed referrers this epoch, and how many buyers did they bring?”
  • antseedstats_referrals_detail/referrals/{address}
    “What is the referral rank of 0xa193...25ac and what has it earned from referrals?”
  • antseedstats_referrals_builders/referrals/builders
    “Which apps route the most usage to Antseed and what do they earn?”
  • antseedstats_diem/diem
    “How much DIEM is staked on Antseed and what has the pool paid out?”
  • antseedstats_diem_proxy/diem/proxy
    “What has the Venice.ai proxy seller earned, and how much DIEM is queued to leave its pool?”
  • antseedstats_epochs/epochs
    “Show every Antseed epoch with its emission, points and burn”
  • antseedstats_epochs_detail/epochs/{n}
    “What is the projected $ANTS per USD for buyers in the current epoch?”
  • antseedstats_ants_supply/ants/supply
    “How much $ANTS has been minted, burned and staked so far?”
  • antseedstats_ants_overview/ants
    “How much $ANTS exists, how much was burned each epoch, and who holds it?”
  • antseedstats_ants_holders/ants/holders
    “Who holds the most $ANTS, and how many holders are there?”
  • antseedstats_channels/channels
    “How many payment channels are open on Antseed, and which were active last?”
  • antseedstats_channels_detail/channels/{channelId}
    “Show channel 0x...0001: buyer, seller, what it settled and which models”
  • antseedstats_requests/requests
    “What were the last 20 inference requests settled on Antseed?”
  • antseedstats_transactions/transactions
    “What happened on Antseed's contracts in the last few minutes?”
  • antseedstats_search/search
    “Find the seller called Venice.ai Proxy and the model id for Kimi K3”
  • antseedstats_estimate/estimate
    “What would 22M input and 5.5M output tokens of Kimi K3 cost on Antseed vs list?”
  • antseedstats_glossary/glossary
    “How does AntSeedStats define network revenue and organic?”
  • antseedstats_changelog/changelog
    “What changed on AntSeedStats this week?”
  • antseedstats_usage/usage
    “How many calls does the AntSeedStats API get, and from what kind of clients?”
  • antseedstats_status/status
    “Is the AntSeedStats indexer up to date with the Base chain?”
Paid answers (x402)
POST/askAsk a question, $0.03 in USDC on Base

A question in plain language, answered from this site's own data with the pages and API calls it used. Paid per answer with x402: USDC on Base, no account, no ETH (the facilitator pays the gas). Paid to 0x1B3eDb34308C17D43beB5e3631AadAD54d813dd0.

Without payment: a 402 with the price in PAYMENT-REQUIRED
curl -i -X POST https://antseedstats.com/api/v1/ask \
  -H 'content-type: application/json' \
  -d '{"question":"Cheapest trusted seller for Kimi K3?"}'
Paying with an x402 client (Node)
import { wrapFetchWithPaymentFromConfig } from "@x402/fetch";
import { ExactEvmScheme } from "@x402/evm/exact/client";
import { privateKeyToAccount } from "viem/accounts";

const pay = wrapFetchWithPaymentFromConfig(fetch, {
  schemes: [{ network: "eip155:8453", client: new ExactEvmScheme(privateKeyToAccount(process.env.PK)) }],
});
const r = await pay("https://antseedstats.com/api/v1/ask", {
  method: "POST", headers: { "content-type": "application/json" },
  body: JSON.stringify({ question: "Cheapest trusted seller for Kimi K3?" }),
});
console.log((await r.json()).data.answer);
  • You pay only for an answer: the payment settles only when the response status is 200. A 400 (bad question), 422 (answered without any of our data, such as an off-topic question), 429, 502 (engine failed) or 503 (daily budget reached) is never charged.
  • The same question within 10 minutes returns the stored answer and is charged like a new one: the price buys the answer, not the computation.
  • Payments are final on-chain. A refund happens only when the fault is ours, by hand, to the paying address; ask @gekko_eth on X with the transaction hash.
  • Answers come from the AntSeedStats tools only (the public API), by Claude Haiku 5.5 on Venice with GLM 5.3 Flash as fallback. They cite the pages and API calls they used. The data stays under CC BY 4.0.
  • 10 questions per minute per IP; a daily budget of $5 of inference, past which the answer is a 503 until 00:00 UTC, never charged.

The terms as JSON: GET /api/v1/ask.

REST API
Network
GET/network/healthServices sellers report failing now

Seller-reported service health across the network (metadata v13/v14, SDK 0.2.157+): a seller's node checks each model it lists about every 5 minutes (a 1-token probe, or what live buyer requests showed) and signs the result into its announcement; buyer nodes skip a model marked failing. We read the announcements, we probe no one. Sellers on an older SDK report nothing and are not counted: null is not healthy. by_seller lists every reporting seller, the all-healthy first (best reputation first) then the troubled from the fewest problems to the most; failing_services every service marked failing, with the reason in the SDK's code and in plain words. Figures refresh with each crawl and each 5-minute price re-read.

curl "https://antseedstats.com/api/v1/network/health"
/api/v1
Response fields (50)
  • at_tsinteger
    Unix seconds this view was assembled
  • servicesinteger
    Every advertised service on the network still served (a service failing for a day or more is not served and is out of every figure but not_served)
  • reportedinteger
    Services with a health report
  • failinginteger
    Reported failing, under a day
  • healthyinteger
    Reported healthy
  • unknowninteger
    Reported but not checked yet
  • not_servedinteger
    Failing for a day or more: the seller stopped serving them; listed in not_served_services for the record
  • sellersinteger
    Announcing peers
  • sellers_reportinginteger
    Peers whose announcement carries health
  • sellers_failinginteger
    Reporting peers with at least one failing service
  • performance_servicesinteger
    Services with a last-hour performance report
  • requests_last_hourinteger
    Sum of reported buyer requests over the last hour
  • by_seller[]array
  • by_seller[].addressstring
    Lowercase 0x
  • by_seller[].namestring | null
    Announced name (untrusted)
  • by_seller[].agent_idstring | null
    On-chain agent id; null for an announce-only peer
  • by_seller[].provenboolean
    Has settled on-chain
  • by_seller[].reputationnumber | null
    The SDK's buyer trust score 0 to 100, from the announcement
  • by_seller[].servicesinteger
    Advertised services
  • by_seller[].reportedinteger
    With a health report
  • by_seller[].failinginteger
    Failing
  • by_seller[].healthyinteger
    Healthy
  • by_seller[].unknowninteger
    Not checked yet
  • by_seller[].top_reasonstring | null
    Most common failing reason code
  • by_seller[].top_reason_labelstring | null
    The same in plain words
  • by_seller[].since_tsinteger | null
    Unix seconds of the oldest failure
  • by_seller[].checked_tsinteger | null
    Unix seconds of the newest check
  • by_seller[].requests_last_hourinteger | null
    Reported buyer requests, summed
  • failing_services[]array
  • failing_services[].sellerstring
    Lowercase 0x
  • failing_services[].seller_namestring | null
    Announced name (untrusted)
  • failing_services[].advertised_asstring
    Raw model name
  • failing_services[].reasonstring | null
    SDK reason code
  • failing_services[].reason_labelstring | null
    Plain words
  • failing_services[].http_statusinteger | null
    Upstream HTTP status of the last failed check
  • failing_services[].sourcestring | null
    probe or traffic
  • failing_services[].since_tsinteger | null
    Unix seconds of the first failure
  • failing_services[].checked_tsinteger | null
    Unix seconds of the last check
  • failing_services[].requests_last_hourinteger | null
    Reported buyer requests for it
  • not_served_services[]array
  • not_served_services[].sellerstring
    Lowercase 0x
  • not_served_services[].seller_namestring | null
    Announced name (untrusted)
  • not_served_services[].advertised_asstring
    Raw model name
  • not_served_services[].reasonstring | null
    SDK reason code
  • not_served_services[].reason_labelstring | null
    Plain words
  • not_served_services[].http_statusinteger | null
    Upstream HTTP status of the last failed check
  • not_served_services[].sourcestring | null
    probe or traffic
  • not_served_services[].since_tsinteger | null
    Unix seconds of the first failure, a day or more ago
  • not_served_services[].checked_tsinteger | null
    Unix seconds of the last check
  • not_served_services[].requests_last_hourinteger | null
    Reported buyer requests for it

Cache 90 s. Sources: health = announce, requests_last_hour = announce.

GET/network/summaryLifetime totals, windows, current epoch

The home page numbers. Lifetime settled USDC and fees come from every ChannelSettled event. The 24h, 7d and 30d windows are rolling (not calendar) and count distinct buyers and sellers inside the window. paid_usage_24h comes from settlements whose buyers shared a per-model split, so it is a lower bound.

curl "https://antseedstats.com/api/v1/network/summary"
/api/v1
Response fields (70)
  • lifetime.settled_microinteger
    settled in micro-USDC (integer, 6 decimals)
  • lifetime.settled_usdstring
    settled in USDC, formatted with 2 decimals
  • lifetime.fees_microinteger
    fees in micro-USDC (integer, 6 decimals)
  • lifetime.fees_usdstring
    fees in USDC, formatted with 2 decimals
  • lifetime.settlesinteger
    Settlement events
  • lifetime.buyersinteger
    Distinct buyers that settled
  • lifetime.sellersinteger
    Distinct sellers that settled
  • lifetime.tokens_ininteger
    Input tokens (MetadataRecorded)
  • lifetime.tokens_outinteger
    Output tokens
  • lifetime.requestsinteger
    Requests
  • windowsobject
    Rolling windows ending now
  • windows.24h.settled_microinteger
    settled in micro-USDC (integer, 6 decimals)
  • windows.24h.settled_usdstring
    settled in USDC, formatted with 2 decimals
  • windows.24h.settlesinteger
    Settles
  • windows.24h.buyersinteger
    Distinct buyers
  • windows.24h.sellersinteger
    Distinct sellers
  • windows.7d.settled_microinteger
    settled in micro-USDC (integer, 6 decimals)
  • windows.7d.settled_usdstring
    settled in USDC, formatted with 2 decimals
  • windows.7d.settlesinteger
    Settles
  • windows.7d.buyersinteger
    Distinct buyers
  • windows.7d.sellersinteger
    Distinct sellers
  • windows.30d.settled_microinteger
    settled in micro-USDC (integer, 6 decimals)
  • windows.30d.settled_usdstring
    settled in USDC, formatted with 2 decimals
  • windows.30d.settlesinteger
    Settles
  • windows.30d.buyersinteger
    Distinct buyers
  • windows.30d.sellersinteger
    Distinct sellers
  • paid_usage_24hobject
    Rolling 24h from per-model usage (lower bound)
  • paid_usage_24h.requestsinteger
    Paid requests
  • paid_usage_24h.tokensinteger
    Paid tokens
  • paid_usage_24h.imagesinteger
    Generated images
  • paid_usage_24h.video_secondsinteger
    Seconds of video
  • paid_usage_24h.videosinteger
    Videos delivered
  • paid_usage_24h.free_requestsinteger
    Free-tier requests
  • paid_usage_24h.free_tokensinteger
    Free-tier tokens
  • paid_usage_24h.modelsinteger
    Distinct models used
  • paid_usage_24h.coveragenumber
    Share of settlements carrying a per-model split, 0 to 1
  • generative_7dobject
    Images and video of the last 7 days (lower bound); the home panel
  • generative_7d.imagesinteger
    Generated images
  • generative_7d.videosinteger
    Videos delivered
  • generative_7d.video_secondsinteger
    Seconds of video
  • generative_7d.spend_microinteger
    spend in micro-USDC (integer, 6 decimals)
  • generative_7d.spend_usdstring
    spend in USDC, formatted with 2 decimals
  • modality_7dobject
    Spend of the last 7 days by modality; each attributed event counts in exactly one (lower bound: buyers that share the per-model split); the home By modality panel
  • modality_7d.textobject
    Events with no images, video or speech characters, on a model (not a data service)
  • modality_7d.text.spend_microinteger
    spend in micro-USDC (integer, 6 decimals)
  • modality_7d.text.spend_usdstring
    spend in USDC, formatted with 2 decimals
  • modality_7d.text.tokensinteger
    Input plus output tokens of the text events (no image credit)
  • modality_7d.imagesobject
    Events with images and no video
  • modality_7d.images.spend_microinteger
    spend in micro-USDC (integer, 6 decimals)
  • modality_7d.images.spend_usdstring
    spend in USDC, formatted with 2 decimals
  • modality_7d.images.imagesinteger
    Generated images
  • modality_7d.videoobject
    Events with video seconds
  • modality_7d.video.spend_microinteger
    spend in micro-USDC (integer, 6 decimals)
  • modality_7d.video.spend_usdstring
    spend in USDC, formatted with 2 decimals
  • modality_7d.video.videosinteger
    Videos delivered
  • modality_7d.video.video_secondsinteger
    Seconds of video
  • modality_7d.speechobject | null
    Events with speech characters and no images or video; null on a window computed before speech was counted
  • modality_7d.speech.spend_microinteger
    spend in micro-USDC (integer, 6 decimals)
  • modality_7d.speech.spend_usdstring
    spend in USDC, formatted with 2 decimals
  • modality_7d.speech.charactersinteger
    Characters read aloud (metadata v5, since 4 Nov 2026)
  • modality_7d.dataobject | null
    Events on a data service (a seller's own HTTP API billed per completed request, told apart by its service name); null when none was settled in the window
  • modality_7d.data.spend_microinteger
    spend in micro-USDC (integer, 6 decimals)
  • modality_7d.data.spend_usdstring
    spend in USDC, formatted with 2 decimals
  • modality_7d.data.requestsinteger
    Completed requests
  • modality_7d.attributed_microinteger
    attributed in micro-USDC (integer, 6 decimals)
  • modality_7d.attributed_usdstring
    attributed in USDC, formatted with 2 decimals
  • epoch.numberinteger
    Current epoch
  • epoch.starts_tsinteger
    UTC seconds
  • epoch.ends_tsinteger
    UTC seconds
  • epoch.seconds_leftinteger
    Until the next epoch

Cache 45 s. Sources: lifetime = chain, windows = chain, paid_usage_24h = chain, epoch = chain.

GET/network/dailyPer UTC day: settled, users, tokens, media

One row per UTC calendar day with no gaps. Settled and fees from ChannelSettled; tokens and requests from MetadataRecorded (paid); free usage from the free-tier counters; users are distinct accounts that settled that day; images and video (seconds, videos, spend) from the per-model usage reports, paid only, since free-tier reports carry no image or video counter. The image, video and speech cells are null on days before that counter's first report (no reading existed), and 0 after it is a real zero. Today is partial.

curl "https://antseedstats.com/api/v1/network/daily?from=2026-09-30&to=2026-10-06"
/api/v1
Parameters
  • fromquery · date
    First UTC day, YYYY-MM-DD (default: 30 days ago)
  • toquery · date
    Last UTC day, YYYY-MM-DD (default: today)
  • formatquery · json | csv · default json
    json (default) or csv
Response fields (29)
  • fromstring
    First day returned
  • tostring
    Last day returned
  • days[]array
  • days[].daystring
    UTC calendar day
  • days[].settled_microinteger
    settled in micro-USDC (integer, 6 decimals)
  • days[].settled_usdstring
    settled in USDC, formatted with 2 decimals
  • days[].fees_microinteger
    fees in micro-USDC (integer, 6 decimals)
  • days[].fees_usdstring
    fees in USDC, formatted with 2 decimals
  • days[].settlesinteger
    Settlement events
  • days[].usersinteger
    Distinct accounts that settled (buyers plus sellers)
  • days[].new_usersinteger
    Accounts whose first settlement is this day
  • days[].buyersinteger
    Distinct buyers
  • days[].sellersinteger
    Distinct sellers
  • days[].free_only_usersinteger
    Accounts active on the free tier only
  • days[].tokens_ininteger
    Paid input tokens
  • days[].tokens_outinteger
    Paid output tokens
  • days[].requestsinteger
    Paid requests
  • days[].free_tokensinteger
    Free-tier tokens
  • days[].free_requestsinteger
    Free-tier requests
  • days[].imagesinteger | null
    Images generated (metadata v3 and later; null before the first image report)
  • days[].images_spend_microinteger | null
    images spend in micro-USDC (integer, 6 decimals)
  • days[].images_spend_usdstring | null
    images spend in USDC, formatted with 2 decimals
  • days[].video_secondsinteger | null
    Seconds of video generated (metadata v4, since 8 Oct 2026; null before the first video report)
  • days[].videosinteger | null
    Videos delivered
  • days[].video_spend_microinteger | null
    video spend in micro-USDC (integer, 6 decimals)
  • days[].video_spend_usdstring | null
    video spend in USDC, formatted with 2 decimals
  • days[].charactersinteger | null
    Speech characters read aloud (metadata v5, since 4 Nov 2026; null before the first speech report)
  • days[].speech_spend_microinteger | null
    speech spend in micro-USDC (integer, 6 decimals)
  • days[].speech_spend_usdstring | null
    speech spend in USDC, formatted with 2 decimals

Cache 5 to 8 min. Sources: days = chain.

GET/leadersLast 24h leaders, wash left out

The 'Happening now' band. Rolling 24h window. Sellers we treat as wash trading are left out of every ranking (excluded_sellers says how many were active); this is our judgement, not a chain fact. Model rankings use per-model usage shared by buyers, so they are lower bounds.

curl "https://antseedstats.com/api/v1/leaders"
/api/v1
Response fields (33)
  • window_hoursinteger
    24
  • seller.addressstring
    Lowercase 0x
  • seller.namestring | null
    Announced name (untrusted)
  • seller.settled_microinteger
    settled in micro-USDC (integer, 6 decimals)
  • seller.settled_usdstring
    settled in USDC, formatted with 2 decimals
  • seller.buyersinteger
    Distinct buyers
  • top_model_by_sales.modelstring
    Canonical id
  • top_model_by_sales.displaystring
    Name
  • top_model_by_sales.settled_microinteger
    settled in micro-USDC (integer, 6 decimals)
  • top_model_by_sales.settled_usdstring
    settled in USDC, formatted with 2 decimals
  • top_model_by_sales.tokensinteger
    Tokens
  • top_model_by_sales.requestsinteger
    Requests
  • top_model_by_sales.buyersinteger
    Buyers
  • top_model_by_tokens.modelstring
    Canonical id
  • top_model_by_tokens.displaystring
    Name
  • top_model_by_tokens.settled_microinteger
    settled in micro-USDC (integer, 6 decimals)
  • top_model_by_tokens.settled_usdstring
    settled in USDC, formatted with 2 decimals
  • top_model_by_tokens.tokensinteger
    Tokens
  • top_model_by_tokens.requestsinteger
    Requests
  • top_model_by_tokens.buyersinteger
    Buyers
  • top_free_model.modelstring
    Canonical id
  • top_free_model.displaystring
    Name
  • top_free_model.tokensinteger
    Free tokens
  • top_free_model.requestsinteger
    Free requests
  • top_free_model.buyersinteger
    Buyers
  • free_models[]array
    Every model served free in the window, most free tokens first (the home's Top free models card)
  • free_models[].modelstring
    Canonical id
  • free_models[].displaystring
    Name
  • free_models[].tokensinteger
    Free tokens
  • free_models[].requestsinteger
    Free requests
  • free_models[].buyersinteger
    Buyers
  • free_buyersinteger
    Distinct buyers of free usage in the window
  • excluded_sellersinteger
    Wash-trading sellers active in the window and left out

Cache 45 s. Sources: seller = chain, models = chain, exclusion = antseedstats.

GET/distributionWho the volume concentrates on

Lifetime settled USDC split across buyers and across sellers: the top accounts with their share, the concentration indices (largest share, top three, HHI, effective number, Gini) and the Lorenz curve of each side. Every seller counts here, wash included: concentration is a chain fact, not a ranking.

curl "https://antseedstats.com/api/v1/distribution?top=10"
/api/v1
Parameters
  • topquery · int · default 20
    Rows per side
Response fields (36)
  • buyers.rows[]array
  • buyers.rows[].addressstring
    Lowercase 0x
  • buyers.rows[].namestring | null
    Announced name (untrusted)
  • buyers.rows[].spent_microinteger
    spent in micro-USDC (integer, 6 decimals)
  • buyers.rows[].spent_usdstring
    spent in USDC, formatted with 2 decimals
  • buyers.rows[].sharenumber
    Share of the side's total, 0 to 1
  • buyers.rows[].settlesinteger
    Settlements
  • buyers.concentrationobject
    Concentration of settled volume over this side
  • buyers.concentration.totalnumber
    Sum of the weights (USDC settled)
  • buyers.concentration.countinteger
    Non-zero counterparties
  • buyers.concentration.top_sharenumber
    Largest single share, 0 to 1
  • buyers.concentration.top3_sharenumber
    Top three combined, 0 to 1
  • buyers.concentration.hhinumber
    Herfindahl-Hirschman index, sum of squared shares, 0 to 1
  • buyers.concentration.effectivenumber
    1 over HHI: effective number of equal-sized counterparties
  • buyers.concentration.gininumber
    0 even, towards 1 one dominates
  • buyers.lorenz[]array
  • buyers.lorenz[].xnumber
    Share of accounts, 0 to 1
  • buyers.lorenz[].ynumber
    Share of volume, 0 to 1
  • sellers.rows[]array
  • sellers.rows[].addressstring
    Lowercase 0x
  • sellers.rows[].namestring | null
    Announced name (untrusted)
  • sellers.rows[].spent_microinteger
    spent in micro-USDC (integer, 6 decimals)
  • sellers.rows[].spent_usdstring
    spent in USDC, formatted with 2 decimals
  • sellers.rows[].sharenumber
    Share of the side's total, 0 to 1
  • sellers.rows[].settlesinteger
    Settlements
  • sellers.concentrationobject
    Concentration of settled volume over this side
  • sellers.concentration.totalnumber
    Sum of the weights (USDC settled)
  • sellers.concentration.countinteger
    Non-zero counterparties
  • sellers.concentration.top_sharenumber
    Largest single share, 0 to 1
  • sellers.concentration.top3_sharenumber
    Top three combined, 0 to 1
  • sellers.concentration.hhinumber
    Herfindahl-Hirschman index, sum of squared shares, 0 to 1
  • sellers.concentration.effectivenumber
    1 over HHI: effective number of equal-sized counterparties
  • sellers.concentration.gininumber
    0 even, towards 1 one dominates
  • sellers.lorenz[]array
  • sellers.lorenz[].xnumber
    Share of accounts, 0 to 1
  • sellers.lorenz[].ynumber
    Share of volume, 0 to 1

Cache 5 to 8 min. Sources: buyers = chain, sellers = chain, name = announce.

Sellers
GET/sellersEvery seller, lifetime and 6h to 30d

The seller leaderboard. Lifetime figures from ChannelSettled and MetadataRecorded. wash_registry is Antseed's on-chain verdict; organic is our view (registry plus confirmed clusters). The default view is all: nothing is hidden unless you ask.

curl "https://antseedstats.com/api/v1/sellers?sort=revenue_7d&view=organic&limit=10"
/api/v1
Parameters
  • sortquery · revenue_7d | revenue_6h | revenue_24h | revenue_30d | revenue | tokens | requests · default revenue_7d
    Sort key, descending
  • viewquery · all | organic · default all
    organic leaves out sellers in the on-chain wash registry plus our confirmed clusters; all keeps every seller
  • limitquery · int · default 100
    Rows to return (1 to 1000)
  • offsetquery · int · default 0
    Rows to skip
Response fields (36)
  • totalinteger
    Sellers matching the view
  • sellers[]array
  • sellers[].addressstring
    Lowercase 0x address
  • sellers[].agent_idstring | null
    On-chain agent id (null for announce-only peers)
  • sellers[].namestring | null
    Display name from the seller's own announcement (untrusted text)
  • sellers[].verifiedboolean
    Has a domain or GitHub proof we verified
  • sellers[].liveboolean
    Announcing on the DHT right now
  • sellers[].revenue_microinteger
    revenue in micro-USDC (integer, 6 decimals)
  • sellers[].revenue_usdstring
    revenue in USDC, formatted with 2 decimals
  • sellers[].revenue_7d_microinteger
    revenue 7d in micro-USDC (integer, 6 decimals)
  • sellers[].revenue_7d_usdstring
    revenue 7d in USDC, formatted with 2 decimals
  • sellers[].revenue_6h_microinteger
    revenue 6h in micro-USDC (integer, 6 decimals)
  • sellers[].revenue_6h_usdstring
    revenue 6h in USDC, formatted with 2 decimals
  • sellers[].revenue_prev_6h_microinteger
    revenue prev 6h in micro-USDC (integer, 6 decimals)
  • sellers[].revenue_prev_6h_usdstring
    revenue prev 6h in USDC, formatted with 2 decimals
  • sellers[].revenue_24h_microinteger
    revenue 24h in micro-USDC (integer, 6 decimals)
  • sellers[].revenue_24h_usdstring
    revenue 24h in USDC, formatted with 2 decimals
  • sellers[].revenue_prev_24h_microinteger
    revenue prev 24h in micro-USDC (integer, 6 decimals)
  • sellers[].revenue_prev_24h_usdstring
    revenue prev 24h in USDC, formatted with 2 decimals
  • sellers[].revenue_30d_microinteger
    revenue 30d in micro-USDC (integer, 6 decimals)
  • sellers[].revenue_30d_usdstring
    revenue 30d in USDC, formatted with 2 decimals
  • sellers[].revenue_prev_30d_microinteger
    revenue prev 30d in micro-USDC (integer, 6 decimals)
  • sellers[].revenue_prev_30d_usdstring
    revenue prev 30d in USDC, formatted with 2 decimals
  • sellers[].fees_microinteger
    fees in micro-USDC (integer, 6 decimals)
  • sellers[].fees_usdstring
    fees in USDC, formatted with 2 decimals
  • sellers[].settlesinteger
    Lifetime settlement events
  • sellers[].channelsinteger
    Lifetime payment channels
  • sellers[].tokensinteger
    Lifetime tokens served, input plus output (MetadataRecorded)
  • sellers[].requestsinteger
    Lifetime requests (MetadataRecorded)
  • sellers[].modelsinteger
    Models advertised
  • sellers[].reputationnumber | null
    Trust score 0 to 100 (Antseed SDK formula, recomputed by us)
  • sellers[].ghost_ratenumber | null
    Share of channels opened and never settled, 0 to 1
  • sellers[].wash_registryboolean
    Proven wash trader in Antseed's on-chain registry (chain fact)
  • sellers[].organicboolean
    In our Organic view: not in the registry and not in a cluster we confirmed (our judgement)
  • sellers[].first_settle_tsinteger | null
    First settlement, UTC seconds (null if none)
  • sellers[].last_settle_tsinteger | null
    Last settlement, UTC seconds (null if none)

Cache 5 to 8 min. Sources: figures = chain, name = announce, reputation = antseedstats, wash_registry = chain, organic = antseedstats.

GET/sellers/{address}One seller: totals, trust, verification

Everything the seller page shows. trust holds the parts of the Antseed trust score as we recompute them from on-chain inputs with the SDK (0.2.168: power up to 70, usage up to 30, identity up to +20, free model +5, capped at 100, times the model-check multiplier; history is reported but scores 0); wash_registry comes from the registry contract.

curl "https://antseedstats.com/api/v1/sellers/0x1f228613116e2d08014dfdcc198377c8dedf18c9"
/api/v1
Parameters
  • addresspath · address · required
    Seller address (0x, 40 hex, any case)
Response fields (50)
  • addressstring
    Lowercase 0x
  • agent_idstring | null
    On-chain agent id
  • peer_idstring | null
    DHT peer id
  • namestring | null
    Announced name (untrusted)
  • model_checksobject | null
    Antseed's model checks, read live off AntseedVerificationResults; null when the seller has no live verdict or the state could not be read (see the page: never shown as x0 on a failed read)
  • model_checks.sameinteger
    Live SAME verdicts (answers as the advertised model), every service of the agent
  • model_checks.diffinteger
    Live DIFF verdicts (very likely another model; two auditors agreed)
  • model_checks.points_multipliernumber | null
    The verification policy's multiplier on this seller's, its buyers' and its pool's points from each settlement on (rates read from the policy); null when no policy is registered
  • model_checks.read_tsinteger
    UTC seconds of the read (every 5 min)
  • rankinteger | null
    Position by lifetime revenue among the Organic sellers; null for a seller outside Organic (see organic)
  • sharenumber | null
    Share of the Organic sellers' lifetime revenue, 0 to 1; null outside Organic
  • revenue_microinteger
    revenue in micro-USDC (integer, 6 decimals)
  • revenue_usdstring
    revenue in USDC, formatted with 2 decimals
  • revenue_7d_microinteger
    revenue 7d in micro-USDC (integer, 6 decimals)
  • revenue_7d_usdstring
    revenue 7d in USDC, formatted with 2 decimals
  • revenue_prev_7d_microinteger
    revenue prev 7d in micro-USDC (integer, 6 decimals)
  • revenue_prev_7d_usdstring
    revenue prev 7d in USDC, formatted with 2 decimals
  • fees_microinteger
    fees in micro-USDC (integer, 6 decimals)
  • fees_usdstring
    fees in USDC, formatted with 2 decimals
  • settlesinteger
    Settlements
  • channelsinteger
    Channels
  • buyersinteger
    Distinct buyers
  • tokensinteger
    Tokens served
  • requestsinteger
    Requests
  • reputationnumber | null
    Trust score 0 to 100
  • trust.historynumber | null
    0 to 50
  • trust.usagenumber | null
    0 to 20
  • trust.powernumber | null
    0 to 10
  • trust.identitynumber | null
    0 to 20
  • trust.identity_kindstring | null
    Kind of identity claim
  • trust.free_modelsnumber | null
    5 when the seller announces a free model, else null (added on top of the parts, total capped at 100)
  • trust.epochinteger | null
    Epoch the score was computed for
  • ghost_ratenumber | null
    Channels never settled over channels opened, 0 to 1
  • stake_usdcnumber | null
    USDC stake on-chain
  • verifiedboolean
    Owner ✓: has a domain or GitHub proof we verified. About who runs the seller, not about its models (that is model_checks)
  • verification[]array
  • verification[].kindstring
    domain or github
  • verification[].targetstring
    Domain or GitHub user
  • verification[].verifiedboolean
    Effective state
  • verification[].hrefstring
    Verified resource
  • liveboolean
    Announcing on the DHT now
  • wash_registryboolean
    Proven wash trader on-chain
  • organicboolean
    In our Organic view
  • routers[]array
    Routing services it last announced (irp-suggest); empty when not a router. A router picks a seller for the buyer; it is never listed as a model
  • data_services[]array
    Data services it announces (openapi-data, metadata v18): its own HTTP API, billed per completed request. Not models: never in a model list or ranking
  • data_services[].servicestring
    Announced service name
  • data_services[].usd_per_requestnumber | null
    USD per completed request (a 2xx answer with a body); null when no price is announced
  • data_services[].freeboolean
    The request price is 0
  • first_settle_tsinteger | null
    UTC seconds
  • last_settle_tsinteger | null
    UTC seconds

Cache 45 s. Sources: figures = chain, name = announce, prices = announce, reputation = antseedstats, verification = antseedstats, wash_registry = chain, organic = antseedstats.

GET/sellers/{address}/dailyOne seller per UTC day (the cross-check)

The cross-check series. settled is what changed hands on-chain for this seller that day (every ChannelSettled where it is the seller), which is why it can differ from what the seller measured as served. Tokens come from MetadataRecorded for the seller's agent and are null for announce-only peers.

curl "https://antseedstats.com/api/v1/sellers/0x1f228613116e2d08014dfdcc198377c8dedf18c9/daily?from=2026-09-30&to=2026-10-06"
/api/v1
Parameters
  • addresspath · address · required
    Seller address (0x, 40 hex, any case)
  • fromquery · date
    First UTC day, YYYY-MM-DD (default: 30 days ago)
  • toquery · date
    Last UTC day, YYYY-MM-DD (default: today)
  • formatquery · json | csv · default json
    json (default) or csv
Response fields (13)
  • addressstring
    Lowercase 0x
  • fromstring
    First day
  • tostring
    Last day
  • days[]array
  • days[].daystring
    UTC day
  • days[].settled_microinteger
    settled in micro-USDC (integer, 6 decimals)
  • days[].settled_usdstring
    settled in USDC, formatted with 2 decimals
  • days[].fees_microinteger
    fees in micro-USDC (integer, 6 decimals)
  • days[].fees_usdstring
    fees in USDC, formatted with 2 decimals
  • days[].settlesinteger
    Settlements
  • days[].buyersinteger
    Distinct buyers
  • days[].tokens_ininteger | null
    Input tokens
  • days[].tokens_outinteger | null
    Output tokens

Cache 5 to 8 min. Sources: days = chain.

GET/sellers/{address}/checksModel checks of one seller

Auditors approved by Antseed send a seller fixed questions and compare its answers with the real model's; SAME means it answers as the model it advertises, DIFF that it very likely serves another (two auditors must agree). Each check comes with the counts behind it from the auditors' evidence on IPFS, kept only when its hash matches the one on-chain: answers that differed against the real model's own noise. models is each advertised service's latest check, from whichever round audited it last, DIFF first; counts_now says whether it is the service's live verdict on-chain now (a verdict counts until its valid_until_epoch unless a newer audit replaces it; the newer one may count before its evidence is read), so an older fail stays on record without reading as a live one; a PENDING right after it can keep beside it the recorded check that still counts. epoch is the live epoch. rounds is every round with the multiplier right after it. live and points_multiplier are read off the chain every 5 minutes and null when that read is stale (then epoch and counts_now are null too).

curl "https://antseedstats.com/api/v1/sellers/0xd19ffae0e85b5230422019926b2f37ec0f1f41f3/checks"
/api/v1
Parameters
  • addresspath · address · required
    Seller address (0x, 40 hex, any case)
Response fields (88)
  • addressstring
    Lowercase 0x
  • agent_idstring | null
    Agent id
  • namestring | null
    Announced name (untrusted)
  • epochinteger | null
    The live epoch of Antseed's verification results; null when the live state is missing or stale
  • unreadinteger | null
    Live verdicts the contract counts for this seller, recorded in the last 24 hours, that no check here counts yet (a newer audit or a decided PENDING whose evidence is still on its way from IPFS): while above 0, models is behind live. null when the live state is missing or stale
  • evidence_missinginteger | null
    Live verdicts counted for over 24 hours whose evidence never reached us (not published, or its hash did not match): the chain counts them, we cannot show their figures. null when the live state is missing or stale
  • liveobject | null
    Live off the chain; null when not audited or the read is stale
  • live.sameinteger
    Models passing now
  • live.diffinteger
    Models failing now
  • live.points_multipliernumber | null
    The verification policy's multiplier on its $ANTS points now
  • models[]array
    Each advertised service's latest check, whichever round, DIFF first; counts_now marks the ones that count now
  • models[].servicestring
    The name the seller advertises
  • models[].modelstring
    The model as the auditor names it
  • models[].epochinteger
    Epoch of the round
  • models[].audited_tsinteger
    Unix seconds of the latest report
  • models[].first_roundboolean
    The 21-oct-2026 round on Antseed's first audit contract, before checks counted for points
  • models[].verdictstring
    The verdict the contract recorded for that round. PENDING: one DIFF report, a second auditor decides
  • models[].close_callboolean
    A DIFF no auditor could call on its first set of questions (decided after extra sets)
  • models[].not_testedstring | null
    Why no auditor could test it; null when tested
  • models[].valid_until_epochinteger | null
    The check counts while the epoch is below this; null for the first round
  • models[].counts_nowboolean | null
    Whether this check is the service's live verdict on-chain now (from the same live read as live; never the first round). Its SAME or DIFF is what live and the points multiplier are made of; an UNDETERMINED counts for nothing. The page's main table shows these and folds the rest below it. null when our data holds no recorded verdict for this check yet (a PENDING, or a SAME/UNDETERMINED whose status event is not read yet: see valid_until_epoch null) or the live state is missing or stale
  • models[].diffsinteger | null
    Answers that differed in the lead read (lead_auditor's deciding read); reads are never added
  • models[].totalinteger | null
    Answers compared in that read
  • models[].lead_auditorstring | null
    The read the row's figures come from: among the reads that agree with the verdict, the one with the largest margin (furthest over the line for a DIFF, closest to it for a SAME)
  • models[].auditors_splitboolean
    The auditors that got answers disagreed (one DIFF, one SAME); each read is in reads
  • models[].noise_ceilingnumber | null
    The lead read's 99% upper bound of the noise rate
  • models[].fail_lineinteger | null
    The lead read's fail line: this many different answers or more fails
  • models[].line_ratenumber | null
    That line as a share of the read's answers
  • models[].margininteger | null
    diffs − fail_line in the lead read: 0 or more fails, below 0 passes
  • models[].times_noisenumber | null
    The lead read's share that differed ÷ the real model's own share in it
  • models[].noise_diffsinteger | null
    The real model's own differences in the lead read
  • models[].noise_totalinteger | null
    The real model's answers compared in it
  • models[].reads[]array
    Each auditor's read
  • models[].reads[].auditorstring
    Auditor address, lowercase
  • models[].reads[].audited_tsinteger
    Unix seconds the report reached the chain
  • models[].reads[].verdictstring
    This auditor's own verdict
  • models[].reads[].outcomestring | null
    Why the audit could not run (the seller's error code); null when it ran
  • models[].reads[].first_diffsinteger | null
    Answers that differed on the first set of questions
  • models[].reads[].first_totalinteger | null
    Questions answered on the first set
  • models[].reads[].p_valuenumber | null
    Binomial p-value of the first set
  • models[].reads[].extra_setsinteger
    Extra sets of questions run on a close call
  • models[].reads[].diffsinteger | null
    Answers that differed in the deciding read (pooled over the extra sets when they were stored pooled)
  • models[].reads[].totalinteger | null
    Answers in the deciding read
  • models[].reads[].noise_diffsinteger | null
    The real model's answers that differed from each other in the same read: the noise
  • models[].reads[].noise_totalinteger | null
    The real model's answers compared
  • models[].reads[].pooled_p_valuenumber | null
    p-value of the pooled read; null without extra sets
  • models[].reads[].noise_ceilingnumber | null
    The 99% upper bound of the noise rate in the deciding read, what the seller is tested against
  • models[].reads[].proof_urlstring | null
    The seller's evidence file (questions, answers, comparison) through an IPFS gateway
  • models[].reads[].fail_lineinteger | null
    This many different answers or more fails the deciding read: the line the verdict used, from that read's own noise ceiling and the rule's p-value threshold
  • models[].reads[].margininteger | null
    diffs − fail_line: 0 or more fails, below 0 passes
  • models[].reads[].alphanumber | null
    The p-value the deciding read had to fall below to fail: 0.005 on a first set or a look before the last, 0.05 on the last look and in the 20-oct first round; null under a rule we do not know
  • models[].reads[].planned_fail_lineinteger | null
    The first set's line planned before the run, in the seller's evidence file (criticalMismatchCount); the run's own noise can move the real line above it
  • models[].reads[].probe_countinteger | null
    Questions in the first set
  • models[].reads[].powernumber | null
    Chance the first set catches a seller serving another model
  • models[].reads[].detailobject | null
    From the seller's own evidence file, hash-checked; null until read
  • models[].reads[].detail.subjects[]array
  • models[].reads[].detail.subjects[].subjectstring
    Subject of the questions
  • models[].reads[].detail.subjects[].differedinteger
    Answers that differed
  • models[].reads[].detail.subjects[].answeredinteger
    Answers compared
  • models[].reads[].detail.examples[]array
    The misses furthest from the real model
  • models[].reads[].detail.examples[].domainstring
    Subject
  • models[].reads[].detail.examples[].questionstring
    The question
  • models[].reads[].detail.examples[].referencestring | null
    The real model's answer (number or text)
  • models[].reads[].detail.examples[].answerstring | null
    The seller's answer (number or text)
  • models[].reads[].detail.latency_msinteger | null
    Median answer time
  • models[].reads[].detail.audit_cost_usdnumber | null
    What the audit spent buying from the seller
  • models[].reads[].detail.reasoningobject | null
    How the seller handled being asked not to reason (the reference answers without)
  • models[].reads[].detail.reasoning.strategystring | null
    How the auditor finally asked
  • models[].reads[].detail.reasoning.attempts[]array
  • models[].reads[].detail.reasoning.attempts[].strategystring
    A way to switch reasoning off
  • models[].reads[].detail.reasoning.attempts[].reasoning_tokensinteger | null
    Reasoning tokens the seller still used
  • models[].reads[].detail.reasoning.attempts[].valid_answersinteger | null
    Answers it could parse
  • models[].referencesarray | null
    What the auditors compared against
  • models[].references[].modelstring | null
    Reference model id
  • models[].references[].typestring | null
    Where it was called (openrouter)
  • models[].references[].provider_pin[]array
  • models[].references[].overridesobject | null
    Request settings the auditor forced (reasoning effort)
  • models[].model_idstring | null
    Our canonical model id (for /models/{id})
  • models[].model_displaystring | null
    Model name
  • rounds[]array
    Newest first
  • rounds[].epochinteger
    Epoch
  • rounds[].started_tsinteger
    Unix seconds of the round's first report
  • rounds[].first_roundboolean
    Before checks counted for points
  • rounds[].passedinteger
    SAME
  • rounds[].failedinteger
    DIFF
  • rounds[].not_testedinteger
    UNDETERMINED
  • rounds[].pendinginteger
    DIFF with one report, waiting for a second
  • rounds[].points_multipliernumber | null
    The multiplier right after the round; null before points

Cache 90 s. Sources: checks = chain, evidence = chain, multiplier = chain.

GET/sellers/{address}/modelsWhat a seller advertises and sold

advertised is the signed metadata as the network sees it (token prices in USD per 1M tokens; image and video listings carry their unit price, and modality says which applies). sold is decoded from settlements whose buyers shared a per-model split, so it is a lower bound and names come as advertised.

curl "https://antseedstats.com/api/v1/sellers/0x1f228613116e2d08014dfdcc198377c8dedf18c9/models"
/api/v1
Parameters
  • addresspath · address · required
    Seller address (0x, 40 hex, any case)
Response fields (53)
  • addressstring
    Lowercase 0x
  • model_checksobject | null
    Antseed's model checks, read live off AntseedVerificationResults; null when the seller has no live verdict or the state could not be read (see the page: never shown as x0 on a failed read)
  • model_checks.sameinteger
    Live SAME verdicts (answers as the advertised model), every service of the agent
  • model_checks.diffinteger
    Live DIFF verdicts (very likely another model; two auditors agreed)
  • model_checks.points_multipliernumber | null
    The verification policy's multiplier on this seller's, its buyers' and its pool's points from each settlement on (rates read from the policy); null when no policy is registered
  • model_checks.read_tsinteger
    UTC seconds of the read (every 5 min)
  • advertised[]array
  • advertised[].modelstring
    Canonical id
  • advertised[].model_displaystring
    Name
  • advertised[].advertised_asstring
    Raw announced name
  • advertised[].input_usd_per_mnumber | null
    USD per 1M input
  • advertised[].cached_input_usd_per_mnumber | null
    USD per 1M cached input
  • advertised[].output_usd_per_mnumber | null
    USD per 1M output
  • advertised[].freeboolean
    Text: both token rates are 0. Image: the image price is 0. Speech: the character price is 0. Data: the request price is 0. Video: never
  • advertised[].protocolsstring | null
    CSV of API protocols
  • advertised[].categoriesstring | null
    Seller tags, CSV
  • advertised[].list_input_usd_per_mnumber | null
    List USD per 1M input (OpenRouter, or Venice when OpenRouter has none); null when there is none
  • advertised[].list_sourcestring | null
    Where the list price comes from
  • advertised[].save_pctnumber | null
    Percent below list (1-99, 100 when free). Text: input rate vs list_input_usd_per_m. Image: image_usd vs list_image_usd. Speech: usd_per_1m_chars vs list_usd_per_1m_chars. Null when not cheaper, no list price, or a video model
  • advertised[].modalitystring
    What the listing serves; for image, video, speech and data the token prices are protocol fallbacks (null). data is a seller's own HTTP API, billed per completed request: not a model
  • advertised[].image_usdnumber | null
    USD per generated image; null when not unit-billed per image
  • advertised[].video_usd_per_secondnumber | null
    USD per second of video (720p when priced by resolution)
  • advertised[].video_usd_per_videonumber | null
    USD per finished video, for sellers pricing per generation
  • advertised[].list_image_usdnumber | null
    Venice list USD per image (image listings only)
  • advertised[].usd_per_1m_charsnumber | null
    Speech: USD per 1M input characters (the text read aloud); null when not a speech listing
  • advertised[].list_usd_per_1m_charsnumber | null
    Speech: Venice list USD per 1M characters (speech listings only); null when Venice has no list price for the model
  • advertised[].speech_voicesarray | null
    Speech: the voices it accepts; another voice is refused unless speech_custom_voices
  • advertised[].speech_formatsarray | null
    Speech: the response formats it returns
  • advertised[].speech_custom_voicesboolean | null
    Speech: also takes account-specific voice ids
  • advertised[].usd_per_requestnumber | null
    Data: USD per completed request (0 is free); null when not a data listing or no price is announced
  • advertised[].model_checkstring | null
    Antseed's live model check on this listing: SAME (answers as the model) or DIFF (very likely another model); null when none is live
  • advertised[].healthstring | null
    Seller-reported state of this service: healthy, failing or unknown. null when the seller reports none (SDK older than 0.2.157), which is not the same as healthy
  • advertised[].health_reasonstring | null
    Why it fails, the SDK's code: timeout, rate_limited, out_of_credit, auth, model_unavailable, upstream_error, unreachable, init_failed
  • advertised[].health_sourcestring | null
    probe (the seller's own 5-minute check) or traffic (live buyer requests)
  • advertised[].health_checked_tsinteger | null
    Unix seconds of the evidence behind health
  • advertised[].health_since_tsinteger | null
    Unix seconds of the first failure of the current streak
  • advertised[].health_http_statusinteger | null
    Upstream HTTP status of the last failed check
  • advertised[].health_not_servedboolean
    Failing for a day or more: the seller stopped serving it; out of every count and catalog, listed for the record
  • advertised[].requests_last_hourinteger | null
    Buyer requests in the seller's last hour, as the seller counts them
  • advertised[].success_ratenumber | null
    Share of them answered without a server-side failure, 0 to 1
  • advertised[].first_token_ms_p50integer | null
    Median ms to the first streamed token, seller-measured
  • advertised[].first_token_ms_p90integer | null
    90th percentile ms to the first token
  • advertised[].tokens_per_second_p50integer | null
    Median output speed, tokens per second
  • sold[]array
  • sold[].modelstring
    Canonical id
  • sold[].model_displaystring
    Name
  • sold[].settled_microinteger
    settled in micro-USDC (integer, 6 decimals)
  • sold[].settled_usdstring
    settled in USDC, formatted with 2 decimals
  • sold[].tokensinteger
    Tokens
  • sold[].requestsinteger
    Requests
  • sold[].buyersinteger
    Distinct buyers
  • sold[].otherboolean | null
    True only on the catch-all row: usage the Antseed SDK keeps in one bucket once a channel passes 31 models (model "other")
  • sold[].databoolean
    A data service, not a model: model is its announced service name

Cache 90 s. Sources: advertised = announce, sold = chain.

GET/registryAntseed's on-chain wash registry

The sellers Antseed itself has proven to be wash trading, as written in the on-chain registry (M001, 6-sep-2026): the list is a chain fact we only read. Our own clusters are not in it; they stay a flag (organic) on the seller endpoints. Usage points from epoch 22 leave these sellers out.

curl "https://antseedstats.com/api/v1/registry"
/api/v1
Response fields (5)
  • totalinteger
    Sellers in the registry
  • sellers[]array
  • sellers[].addressstring
    Lowercase 0x
  • sellers[].agent_idstring | null
    On-chain agent id
  • sellers[].namestring | null
    Announced name (untrusted)

Cache 45 s. Sources: address = chain, name = announce, agent_id = chain.

Buyers
GET/buyers/topWho is buying now: this week and top

The /buyers page. totals counts every wallet over rolling windows: buyers in the last 24 hours and 7 days, the 7 days before (the page's comparison), USD settled in 7 days and the share of it from the 10 largest buyers, new buyers (first settlement ever in the last 7 days) and what they spent, and regulars (settled in each of the last 4 closed calendar weeks, Monday to Sunday UTC). free_tier counts the wallets on the free tier only this week (buyer-signed, unpaid reports, easier to fake than a payment). spend_mix_7d sorts the week's buyers into spend bands. buyers ranks wallets by USD settled in the period; the organic view counts only spend with organic sellers and leaves out buyers with none. A buyer is a wallet: Antseed has no buyer names.

curl "https://antseedstats.com/api/v1/buyers/top?period=7d&limit=10"
/api/v1
Parameters
  • periodquery · 24h | 7d | 30d · default 7d
    Rolling window of the ranking
  • viewquery · all | organic · default organic
    organic leaves out sellers in the on-chain wash registry plus our confirmed clusters; all keeps every seller
  • limitquery · int · default 100
    Rows to return (1 to 1000)
  • offsetquery · int · default 0
    Rows to skip
Response fields (37)
  • periodstring
    The window ranked
  • viewstring
    organic or all
  • totals.buyers_24hinteger
    Wallets that settled in the last 24 hours
  • totals.buyers_7dinteger
    Wallets that settled in the last 7 days
  • totals.buyers_prev_7dinteger
    Wallets that settled 14 to 7 days ago
  • totals.spent_7d_microinteger
    spent 7d in micro-USDC (integer, 6 decimals)
  • totals.spent_7d_usdstring
    spent 7d in USDC, formatted with 2 decimals
  • totals.top10_share_7dnumber | null
    Share of spent_7d from the week's 10 largest buyers, 0 to 1
  • totals.new_7dinteger
    Wallets whose first settlement ever is in the last 7 days
  • totals.new_7d_spent_microinteger
    new 7d spent in micro-USDC (integer, 6 decimals)
  • totals.new_7d_spent_usdstring
    new 7d spent in USDC, formatted with 2 decimals
  • totals.regularsinteger
    Wallets that settled in each of the last regular_weeks closed calendar weeks
  • totals.regular_weeksinteger
    Closed calendar weeks a regular settled in, each
  • free_tier.only_7dinteger
    Wallets that used the free tier in the last 7 days and settled nothing in them (buyer-signed, unpaid reports)
  • free_tier.new_7dinteger
    Free-only wallets first seen in the last 7 days
  • free_tier.regularsinteger
    Free-only wallets that used the free tier in each of the last regular_weeks closed calendar weeks
  • free_tier.then_paid_7dinteger
    Of the last 7 days' first-time payers, the wallets that used the free tier before
  • spend_mix_7dobject
    The last 7 days' buyers by what each one settled, every seller
  • spend_mix_7d.bands[]array
  • spend_mix_7d.bands[].bandstring
    under1, 1to10, 10to100 or 100plus
  • spend_mix_7d.bands[].from_usdnumber
    Lower bound of the band, USD, inclusive
  • spend_mix_7d.bands[].buyersinteger
    Wallets whose 7-day spend falls in the band
  • spend_mix_7d.bands[].spent_microinteger
    spent in micro-USDC (integer, 6 decimals)
  • spend_mix_7d.bands[].spent_usdstring
    spent in USDC, formatted with 2 decimals
  • spend_mix_7d.one_sellerinteger
    Buyers of the 7 days that paid a single seller
  • spend_mix_7d.five_plus_sellersinteger
    Buyers of the 7 days that paid five sellers or more
  • totalinteger
    Buyers in the period and view
  • buyers[]array
  • buyers[].rankinteger
    Position in the period and view
  • buyers[].addressstring
    Lowercase 0x
  • buyers[].spent_microinteger
    spent in micro-USDC (integer, 6 decimals)
  • buyers[].spent_usdstring
    spent in USDC, formatted with 2 decimals
  • buyers[].sharenumber
    Share of the period's USD in the view, 0 to 1
  • buyers[].sellersinteger
    Distinct sellers paid in the period (organic sellers only in the organic view)
  • buyers[].first_settle_tsinteger
    First settlement ever, UTC seconds
  • buyers[].is_newboolean
    First settlement in the last 7 days
  • buyers[].streak_weeksinteger
    Calendar weeks in a row with a settlement, up to the current week (or the last closed one); 0 under 2

Cache 5 to 8 min. Sources: totals = chain, free_tier = announce, spend_mix_7d = chain, buyers = chain.

GET/buyers/dailyBuyers per UTC day, paid and free

The /buyers chart. active: wallets that settled at least one paid channel that UTC day. new: wallets whose first settlement ever was that day. total: wallets that had settled by the end of the day. The free_ columns are the free tier, from buyer-signed, unpaid reports (easier to fake than a payment): free_active, wallets active that day only through free usage; free_new, wallets first seen on the free tier that day that have never paid; free_total, the running count of those. Every wallet, no gaps; today is partial.

curl "https://antseedstats.com/api/v1/buyers/daily?from=2026-10-01&to=2026-10-31"
/api/v1
Parameters
  • fromquery · date
    First UTC day, YYYY-MM-DD (default: 30 days ago)
  • toquery · date
    Last UTC day, YYYY-MM-DD (default: today)
  • formatquery · json | csv · default json
    json (default) or csv
Response fields (10)
  • fromstring
    First day
  • tostring
    Last day
  • days[]array
  • days[].daystring
    UTC day
  • days[].activeinteger
    Wallets that settled that day
  • days[].newinteger
    Wallets whose first settlement was that day
  • days[].totalinteger
    Wallets that had settled by the end of that day
  • days[].free_activeinteger
    Wallets active that day only through the free tier
  • days[].free_newinteger
    Free-only wallets first seen that day
  • days[].free_totalinteger
    Free-only wallets seen by the end of that day

Cache 5 to 8 min. Sources: days = chain.

GET/buyers/historyBuyers since the start: retention

The /buyers all-time figures. total_ever counts every wallet that settled a paid channel; free_only_wallets the ones that only used the free tier (never in total_ever). still_buying: of the wallets whose first settlement is older than 30 days, the ones that settled in the last 30; the page shows the all view, organic reads only settlements with organic sellers (first and last ones included). mix_7d splits the last 7 days' buyers into new (first settlement ever in the 7 days), regular (settled in each of the last 4 closed calendar weeks) and returning (the rest). weeks are calendar weeks, Monday 00:00 UTC, from the first settlement on; the last one is partial; new is the week of a wallet's first settlement ever, with any seller, in both views. cohorts group wallets by that first week; active[k] counts the members that settled k weeks later.

curl "https://antseedstats.com/api/v1/buyers/history"
/api/v1
Parameters
  • viewquery · all | organic · default all
    organic leaves out sellers in the on-chain wash registry plus our confirmed clusters; all keeps every seller
Response fields (34)
  • viewstring
    all or organic: the view of weeks
  • total_everinteger
    Wallets that ever settled a paid channel
  • free_only_walletsinteger
    Wallets that used only the free tier (buyer-signed reports)
  • free_then_paidinteger
    Wallets that used the free tier before their first payment (in total_ever)
  • still_buying.daysinteger
    The window, in days
  • still_buying.all.baseinteger
    Wallets whose first settlement is older than the window
  • still_buying.all.keptinteger
    Of those, the ones that settled inside it
  • still_buying.all.sharenumber | null
    kept over base, 0 to 1
  • still_buying.organic.baseinteger
    Same, organic sellers only
  • still_buying.organic.keptinteger
    Same, organic sellers only
  • still_buying.organic.sharenumber | null
    kept over base, 0 to 1
  • kept_from_prev_7dinteger
    Of the wallets that settled 14 to 7 days ago, the ones that settled again in the last 7 days
  • mix_7dobject
    The last 7 days' buyers in three groups that add up to the week's buyers, every seller
  • mix_7d.new.buyersinteger
    Wallets
  • mix_7d.new.spent_microinteger
    spent in micro-USDC (integer, 6 decimals)
  • mix_7d.new.spent_usdstring
    spent in USDC, formatted with 2 decimals
  • mix_7d.regular.buyersinteger
    Wallets
  • mix_7d.regular.spent_microinteger
    spent in micro-USDC (integer, 6 decimals)
  • mix_7d.regular.spent_usdstring
    spent in USDC, formatted with 2 decimals
  • mix_7d.returning.buyersinteger
    Wallets
  • mix_7d.returning.spent_microinteger
    spent in micro-USDC (integer, 6 decimals)
  • mix_7d.returning.spent_usdstring
    spent in USDC, formatted with 2 decimals
  • weeks[]array
  • weeks[].weekstring
    Monday of the calendar week, UTC
  • weeks[].buyersinteger
    Wallets that settled in it
  • weeks[].newinteger
    Of those, first settlement ever in it
  • weeks[].returninginteger
    The others
  • weeks[].spent_microinteger
    spent in micro-USDC (integer, 6 decimals)
  • weeks[].spent_usdstring
    spent in USDC, formatted with 2 decimals
  • weeks[].partialboolean
    The current week, still running
  • cohorts[]array
  • cohorts[].weekstring
    Monday of the cohort's first week
  • cohorts[].buyersinteger
    Wallets whose first settlement was in it
  • cohorts[].active[]array
    From k = 0, the cohort's own week

Cache 5 to 8 min. Sources: total_ever = chain, free_only_wallets = announce, free_then_paid = announce, still_buying = chain, weeks = chain, cohorts = chain.

GET/buyers/free-to-paidFrom free to paid

The /buyers section on the free tier's paying wallets. converted counts wallets whose first free use came before their first payment (paid_first, the other way round, is not in it); paid_within says how many paid within a day, a week, a month of that first free use. spent is what they have settled since, spent_7d in the last 7 days, and share_of_spend_7d its share of every buyer's last 7 days. same_seller: converted wallets that have paid the seller of their first free use. weeks groups the wallets that were free when they arrived by the calendar week of their first free use (Monday, UTC) and counts the ones that paid within 7 days; partial while some have not had their 7 days. sellers lists the seller of each wallet's first free use, with at least one conversion. Free usage is signed by the buyer and unpaid (easier to fake); the payments are on-chain settlements.

curl "https://antseedstats.com/api/v1/buyers/free-to-paid"
/api/v1
Response fields (30)
  • convertedinteger
    Wallets whose first free use came before their first payment
  • converted_7dinteger
    Of those, the ones whose first payment fell in the last 7 days
  • paid_within.dayinteger
    First payment within 24 hours of the first free use
  • paid_within.weekinteger
    Within 7 days
  • paid_within.monthinteger
    Within 30 days
  • median_hoursnumber | null
    Median hours from first free use to first payment
  • spent_microinteger
    spent in micro-USDC (integer, 6 decimals)
  • spent_usdstring
    spent in USDC, formatted with 2 decimals
  • spent_7d_microinteger
    spent 7d in micro-USDC (integer, 6 decimals)
  • spent_7d_usdstring
    spent 7d in USDC, formatted with 2 decimals
  • all_spent_7d_microinteger
    all spent 7d in micro-USDC (integer, 6 decimals)
  • all_spent_7d_usdstring
    all spent 7d in USDC, formatted with 2 decimals
  • share_of_spend_7dnumber | null
    spent_7d over all_spent_7d, 0 to 1
  • same_sellerinteger
    Converted wallets that have paid the seller of their first free use
  • paid_firstinteger
    Wallets that paid before their first free use (not in converted)
  • weeks[]array
  • weeks[].weekstring
    Monday of the week of the first free use, UTC
  • weeks[].walletsinteger
    Wallets free when they arrived that week
  • weeks[].paid_within_7dinteger
    Of those, paid within 7 days
  • weeks[].ratenumber
    paid_within_7d over wallets, 0 to 1
  • weeks[].partialboolean
    Some of the week's wallets have not had their 7 days
  • sellers[]array
  • sellers[].addressstring
    Lowercase 0x
  • sellers[].namestring | null
    Announced name (untrusted)
  • sellers[].organicboolean
    Counted by the Organic view
  • sellers[].free_walletsinteger
    Wallets whose first free use was with this seller and had not paid before
  • sellers[].convertedinteger
    Of those, the ones that later paid, any seller
  • sellers[].ratenumber
    converted over free_wallets, 0 to 1
  • sellers[].spent_microinteger
    spent in micro-USDC (integer, 6 decimals)
  • sellers[].spent_usdstring
    spent in USDC, formatted with 2 decimals

Cache 5 to 8 min. Sources: converted = announce, spent = chain, weeks = announce, sellers = announce.

GET/buyersEvery buyer, lifetime spend

The /buyers/all leaderboard. Lifetime spend from every ChannelSettled where the address is the buyer; tokens and requests from MetadataRecorded. spent_organic and sellers_organic count only organic sellers; view=organic lists the buyers with some, and the spent sort then ranks by spent_organic. totals adds the lifetime counts and the escrow flows (also on /buyers/escrow). ants_earned and staked_ants are the $ANTS side (per epoch and per pool in /buyers/{address}/rewards). Free-tier-only buyers are not listed; see /buyers/{address}.

curl "https://antseedstats.com/api/v1/buyers?sort=spent&limit=10"
/api/v1
Parameters
  • viewquery · all | organic · default all
    organic leaves out sellers in the on-chain wash registry plus our confirmed clusters; all keeps every seller
  • sortquery · spent | spent_7d | settles | tokens | requests | last_settle | ants_earned | staked · default spent
    Sort key, descending; spent_7d ranks by the rolling last 7 days
  • limitquery · int · default 100
    Rows to return (1 to 1000)
  • offsetquery · int · default 0
    Rows to skip
Response fields (31)
  • totalinteger
    Buyers listed
  • totals.payinginteger
    Buyers that settled at least once
  • totals.depositorsinteger
    Addresses that deposited into escrow
  • totals.spent_microinteger
    spent in micro-USDC (integer, 6 decimals)
  • totals.spent_usdstring
    spent in USDC, formatted with 2 decimals
  • totals.escrow_deposited_microinteger
    escrow deposited in micro-USDC (integer, 6 decimals)
  • totals.escrow_deposited_usdstring
    escrow deposited in USDC, formatted with 2 decimals
  • totals.escrow_withdrawn_microinteger
    escrow withdrawn in micro-USDC (integer, 6 decimals)
  • totals.escrow_withdrawn_usdstring
    escrow withdrawn in USDC, formatted with 2 decimals
  • totals.escrow_net_microinteger
    escrow net in micro-USDC (integer, 6 decimals)
  • totals.escrow_net_usdstring
    escrow net in USDC, formatted with 2 decimals
  • buyers[]array
  • buyers[].addressstring
    Lowercase 0x address
  • buyers[].spent_microinteger
    spent in micro-USDC (integer, 6 decimals)
  • buyers[].spent_usdstring
    spent in USDC, formatted with 2 decimals
  • buyers[].spent_7d_microinteger | null
    USDC settled in the rolling last 7 days, micro-USDC (null on the detail endpoint)
  • buyers[].spent_7d_usdstring | null
    USDC settled in the rolling last 7 days, 2 decimals (null on the detail endpoint)
  • buyers[].settlesinteger
    Lifetime settlement events as the buyer
  • buyers[].channelsinteger
    Payment channels opened
  • buyers[].sellersinteger
    Distinct sellers paid
  • buyers[].tokensinteger
    Lifetime tokens bought, input plus output (MetadataRecorded)
  • buyers[].requestsinteger
    Lifetime requests (MetadataRecorded)
  • buyers[].deposited_microinteger
    deposited in micro-USDC (integer, 6 decimals)
  • buyers[].deposited_usdstring
    deposited in USDC, formatted with 2 decimals
  • buyers[].first_settle_tsinteger | null
    First settlement, UTC seconds (null if none)
  • buyers[].last_settle_tsinteger | null
    Last settlement, UTC seconds (null if none)
  • buyers[].spent_organic_microinteger
    spent organic in micro-USDC (integer, 6 decimals)
  • buyers[].spent_organic_usdstring
    spent organic in USDC, formatted with 2 decimals
  • buyers[].sellers_organicinteger
    Distinct organic sellers paid
  • buyers[].ants_earnednumber | null
    $ANTS earned in closed epochs since 22 (usage-rewards contract); null when it could not be read
  • buyers[].staked_antsnumber | null
    $ANTS principal in seller pools, counting now or from the next epoch; null when the pools could not be read

Cache 5 to 8 min. Sources: figures = chain, totals = chain.

GET/buyers/escrowBuyer escrow: deposited, withdrawn, net

The /buyers/escrow page. Buyers fund an escrow (AntseedDeposits) before they open channels. deposited sums every deposit, withdrawn every WithdrawalExecuted, net is the difference (an accounting net, not a balance read). depositors counts the wallets that deposited at least once, paying those that settled at least once.

curl "https://antseedstats.com/api/v1/buyers/escrow"
/api/v1
Response fields (8)
  • depositorsinteger
    Wallets that deposited into escrow
  • payinginteger
    Wallets that settled at least once
  • deposited_microinteger
    deposited in micro-USDC (integer, 6 decimals)
  • deposited_usdstring
    deposited in USDC, formatted with 2 decimals
  • withdrawn_microinteger
    withdrawn in micro-USDC (integer, 6 decimals)
  • withdrawn_usdstring
    withdrawn in USDC, formatted with 2 decimals
  • net_microinteger
    net in micro-USDC (integer, 6 decimals)
  • net_usdstring
    net in USDC, formatted with 2 decimals

Cache 5 to 8 min. Sources: figures = chain.

GET/buyers/{address}One buyer: spend, sellers, models

Everything the buyer page shows. sellers is where the spend went; models is the per-model split the buyer itself shared with its settlements (shares_per_model false means it never did, so models is empty). free is the buyer's own free-tier usage reports, never counted as spend. A free-only buyer answers with paid figures at zero.

curl "https://antseedstats.com/api/v1/buyers/0x3f1dbd5c47c6ad8a4b1bb5fb3a3bd8f6e6c5f0f5"
/api/v1
Parameters
  • addresspath · address · required
    Buyer address (0x, 40 hex, any case)
Response fields (46)
  • addressstring
    Lowercase 0x address
  • spent_microinteger
    spent in micro-USDC (integer, 6 decimals)
  • spent_usdstring
    spent in USDC, formatted with 2 decimals
  • spent_7d_microinteger | null
    USDC settled in the rolling last 7 days, micro-USDC (null on the detail endpoint)
  • spent_7d_usdstring | null
    USDC settled in the rolling last 7 days, 2 decimals (null on the detail endpoint)
  • settlesinteger
    Lifetime settlement events as the buyer
  • channelsinteger
    Payment channels opened
  • sellers[]array
  • sellers[].addressstring
    Lowercase 0x
  • sellers[].namestring | null
    Announced name (untrusted)
  • sellers[].spent_microinteger
    spent in micro-USDC (integer, 6 decimals)
  • sellers[].spent_usdstring
    spent in USDC, formatted with 2 decimals
  • sellers[].sharenumber
    Share of this buyer's spend, 0 to 1
  • sellers[].settlesinteger
    Settlements
  • tokensinteger
    Lifetime tokens bought, input plus output (MetadataRecorded)
  • requestsinteger
    Lifetime requests (MetadataRecorded)
  • deposited_microinteger
    deposited in micro-USDC (integer, 6 decimals)
  • deposited_usdstring
    deposited in USDC, formatted with 2 decimals
  • first_settle_tsinteger | null
    First settlement, UTC seconds (null if none)
  • last_settle_tsinteger | null
    Last settlement, UTC seconds (null if none)
  • paidboolean
    Has settled at least one paid channel
  • shares_per_modelboolean
    Its settlements carry a per-model split
  • models[]array
  • models[].modelstring
    Canonical id
  • models[].model_displaystring
    Name
  • models[].spent_microinteger
    spent in micro-USDC (integer, 6 decimals)
  • models[].spent_usdstring
    spent in USDC, formatted with 2 decimals
  • models[].tokensinteger
    Tokens
  • models[].requestsinteger
    Requests
  • models[].sellersinteger
    Sellers it bought this model from
  • models[].otherboolean | null
    True only on the catch-all row: usage the Antseed SDK keeps in one bucket once a channel passes 31 models (model "other")
  • models[].databoolean
    A data service (a seller's own HTTP API billed per completed request), not a model: model is its announced service name
  • freeobject
    Null when the address never reported free usage
  • free.tokensinteger
    Free-tier tokens
  • free.requestsinteger
    Free-tier requests
  • free.sellersinteger
    Distinct sellers
  • free.channelsinteger
    Channels
  • free.first_tsinteger | null
    UTC seconds
  • free.last_tsinteger | null
    UTC seconds
  • free.models[]array
  • free.models[].modelstring
    Canonical id
  • free.models[].model_displaystring
    Name
  • free.models[].tokensinteger
    Tokens
  • free.models[].requestsinteger
    Requests
  • free.models[].sellersinteger
    Sellers
  • free.models[].databoolean
    A data service (a seller's own HTTP API billed per completed request), not a model: model is its announced service name

Cache 45 s. Sources: figures = chain, sellers = chain, models = chain, free = announce, name = announce.

GET/buyers/{address}/dailyOne buyer per UTC day

Spend per UTC calendar day: every ChannelSettled where the address is the buyer, bucketed by day, no gaps from the first settlement to today. Today is partial.

curl "https://antseedstats.com/api/v1/buyers/0x3f1dbd5c47c6ad8a4b1bb5fb3a3bd8f6e6c5f0f5/daily?from=2026-09-30&to=2026-10-06"
/api/v1
Parameters
  • addresspath · address · required
    Buyer address (0x, 40 hex, any case)
  • fromquery · date
    First UTC day, YYYY-MM-DD (default: 30 days ago)
  • toquery · date
    Last UTC day, YYYY-MM-DD (default: today)
  • formatquery · json | csv · default json
    json (default) or csv
Response fields (7)
  • addressstring
    Lowercase 0x
  • fromstring
    First day
  • tostring
    Last day
  • days[]array
  • days[].daystring
    UTC day
  • days[].spent_microinteger
    spent in micro-USDC (integer, 6 decimals)
  • days[].spent_usdstring
    spent in USDC, formatted with 2 decimals

Cache 5 to 8 min. Sources: days = chain.

GET/buyers/{address}/rewardsOne buyer's $ANTS: earned, claimed, staked

The buyer page's Rewards & stake section. Per epoch since 22, what the address's settled usage earned from the usage-rewards contract (pendingBuyerReward, which keeps the amount after a claim) and whether it was claimed (buyerEpochClaimed); the open epoch is what the usage so far would pay and moves until it closes. stakes are the positions it holds in seller pools that count now or from the next epoch, with what it ever settled with that seller: 0 means it stakes behind a seller it never bought from. earned_ants and the epoch figures are null when the contract could not be read; stakes_read false means the pools could not be read, so stakes is unknown, not empty.

curl "https://antseedstats.com/api/v1/buyers/0xa193c4005ede4705ea6b1582bfb45af1907e25ac/rewards"
/api/v1
Parameters
  • addresspath · address · required
    Buyer address (0x, 40 hex, any case)
Response fields (27)
  • addressstring
    Lowercase 0x
  • epochinteger
    Current epoch
  • earned_antsnumber | null
    Earned in closed epochs since 22
  • claimed_antsnumber | null
    Of that, claimed
  • unclaimed_antsnumber | null
    Of that, still claimable
  • accruing_antsnumber | null
    The open epoch so far
  • staked_antsnumber
    Principal in seller pools, counting now or from the next epoch
  • stakes_readboolean
    The pools could be read
  • legacy_claimed_antsnumber
    Claimed from the pre-epoch-22 emissions contract
  • epochs[]array
  • epochs[].epochinteger
    Epoch
  • epochs[].usage_usdnumber
    Settled usage the contract counted, USD
  • epochs[].earned_antsnumber
    ANTS earned
  • epochs[].claimedboolean
    Claimed
  • epochs[].openboolean
    The current epoch (an estimate until it closes)
  • stakes[]array
  • stakes[].position_idinteger
    Position id (lANTS NFT id)
  • stakes[].agent_idinteger
    Pool agent id
  • stakes[].sellerstring | null
    The pool's seller, lowercase 0x
  • stakes[].seller_namestring | null
    Announced name (untrusted)
  • stakes[].principal_antsnumber
    ANTS locked
  • stakes[].epochs_remaininginteger
    Epochs left on the lock
  • stakes[].max_lockedboolean
    At the maximum lock
  • stakes[].pendingboolean
    Counts from the next epoch
  • stakes[].leavingboolean
    Stops counting at the end of this epoch
  • stakes[].spent_with_seller_microinteger
    spent with seller in micro-USDC (integer, 6 decimals)
  • stakes[].spent_with_seller_usdstring
    spent with seller in USDC, formatted with 2 decimals

Cache 45 s. Sources: epochs = chain, stakes = chain, legacy_claimed_ants = chain.

Models and prices
GET/modelsCatalogue: sellers, prices, usage

One row per canonical model (name variants merged). Prices are the min and max advertised across sellers. Usage is lifetime and a lower bound (per-model splits are opt-in). all counts every seller; organic leaves out wash-trading sellers. The default view is organic, like the site; models with no organic usage are still listed.

curl "https://antseedstats.com/api/v1/models?view=organic&limit=10"
/api/v1
Parameters
  • viewquery · all | organic · default organic
    organic leaves out sellers in the on-chain wash registry plus our confirmed clusters; all keeps every seller
  • periodquery · 24h | 7d | 30d | 90d | all · default all
    Adds usage_period: sales, tokens and buyers over the last 24 hours, 7, 30 or 90 days (rolling) in the chosen view, and revenue/tokens sort on it (the page's Top models periods)
  • sortquery · revenue | tokens | requests | sellers · default revenue
    Sort key, descending, applied to the chosen view
  • limitquery · int · default 100
    Rows to return (1 to 1000)
  • offsetquery · int · default 0
    Rows to skip
Response fields (54)
  • totalinteger
    Models
  • top_todayobject | null
    The page's Top sales and Top free boxes (rolling 24 hours, lower bound). null when it could not be read
  • top_today.sales[]array
    The 3 models with the most USDC settled in the last 24 hours, Organic sellers
  • top_today.sales[].modelstring
    Canonical id
  • top_today.sales[].displaystring
    Name
  • top_today.sales[].settled_microinteger
    settled in micro-USDC (integer, 6 decimals)
  • top_today.sales[].settled_usdstring
    settled in USDC, formatted with 2 decimals
  • top_today.sales[].buyersinteger
    Buyers
  • top_today.freearray | null
    The 3 free models with the most tokens served in the last 24 hours, Organic sellers, routers left out; null right after a deploy
  • top_today.free[].modelstring
    Canonical id
  • top_today.free[].displaystring
    Name
  • top_today.free[].free_tokensinteger
    Free-tier tokens
  • top_today.free[].buyersinteger
    Buyers
  • models[]array
  • models[].modelstring
    Canonical id
  • models[].displaystring
    Name
  • models[].customboolean
    A seller's own model, not in any catalog
  • models[].sellersinteger
    Sellers advertising it
  • models[].image_generationboolean
    Priced per image
  • models[].video_generationboolean
    Native video, priced per second or per video
  • models[].min_input_usd_per_mnumber | null
    Cheapest input price
  • models[].max_input_usd_per_mnumber | null
    Dearest input price
  • models[].min_output_usd_per_mnumber | null
    Cheapest output
  • models[].max_output_usd_per_mnumber | null
    Dearest output
  • models[].min_image_usdnumber | null
    Cheapest price per image
  • models[].min_video_usd_per_snumber | null
    Cheapest price per second of video, 720p
  • models[].min_video_usdnumber | null
    Cheapest price per video
  • models[].speech_generationboolean
    Speech (text to audio), priced per character
  • models[].min_usd_per_1m_charsnumber | null
    Cheapest price per 1M characters of speech
  • models[].usageobject
    In the chosen view
  • models[].usage.settled_microinteger
    settled in micro-USDC (integer, 6 decimals)
  • models[].usage.settled_usdstring
    settled in USDC, formatted with 2 decimals
  • models[].usage.tokensinteger
    Tokens
  • models[].usage.imagesinteger
    Images
  • models[].usage.video_secondsinteger
    Seconds of video
  • models[].usage.videosinteger
    Videos delivered
  • models[].usage.requestsinteger
    Requests
  • models[].usage.buyersinteger
    Buyers
  • models[].usage.sellersinteger
    Sellers with sales
  • models[].usage_allobject
    Every seller, always
  • models[].usage_all.settled_microinteger
    settled in micro-USDC (integer, 6 decimals)
  • models[].usage_all.settled_usdstring
    settled in USDC, formatted with 2 decimals
  • models[].usage_all.tokensinteger
    Tokens
  • models[].usage_all.imagesinteger
    Images
  • models[].usage_all.video_secondsinteger
    Seconds of video
  • models[].usage_all.videosinteger
    Videos delivered
  • models[].usage_all.requestsinteger
    Requests
  • models[].usage_all.buyersinteger
    Buyers
  • models[].usage_all.sellersinteger
    Sellers with sales
  • models[].usage_periodobject | null
    Only with period other than all: the window in the chosen view
  • models[].usage_period.settled_microinteger
    settled in micro-USDC (integer, 6 decimals)
  • models[].usage_period.settled_usdstring
    settled in USDC, formatted with 2 decimals
  • models[].usage_period.tokensinteger
    Tokens
  • models[].usage_period.buyersinteger
    Buyers

Cache 5 to 8 min. Sources: prices = announce, usage = chain, organic = antseedstats.

GET/models/{id}/checksModel checks: who passed for one model

Every seller of the model that the approved auditors checked, with the counts behind each verdict from their evidence on IPFS (kept only when its hash matches the one on-chain): answers that differed from the real model against the real model's own noise. sellers is each seller's latest check of the model, from whichever round audited it last, DIFF first; the page's table shows the ones that count now and folds the rest. counts_now says whether it still counts, epoch is the live epoch; every advertised spelling of the model counts. references says what the auditors compared against (model, provider, reasoning setting).

curl "https://antseedstats.com/api/v1/models/claude-fable-5.1/checks"
/api/v1
Parameters
  • idpath · string · required
    Canonical model id or an advertised name
Response fields (84)
  • modelstring
    Canonical id
  • displaystring
    Name
  • services[]array
  • epochinteger | null
    The live epoch of Antseed's verification results; null when the live state is missing or stale
  • unreadinteger | null
    Live verdicts on this model's advertised spellings, recorded in the last 24 hours, that no check here counts yet (evidence still on its way from IPFS): while above 0, the page's figures are behind the chain and it says so. null when the live state is missing or stale
  • evidence_missinginteger | null
    Live verdicts on this model counted for over 24 hours whose evidence never reached us: the chain counts them, we cannot show their figures. null when the live state is missing or stale
  • sellers[]array
    Each seller's latest check, whichever round, DIFF first; counts_now marks the ones that count now
  • sellers[].servicestring
    The name the seller advertises
  • sellers[].modelstring
    The model as the auditor names it
  • sellers[].epochinteger
    Epoch of the round
  • sellers[].audited_tsinteger
    Unix seconds of the latest report
  • sellers[].first_roundboolean
    The 21-oct-2026 round on Antseed's first audit contract, before checks counted for points
  • sellers[].verdictstring
    The verdict the contract recorded for that round. PENDING: one DIFF report, a second auditor decides
  • sellers[].close_callboolean
    A DIFF no auditor could call on its first set of questions (decided after extra sets)
  • sellers[].not_testedstring | null
    Why no auditor could test it; null when tested
  • sellers[].valid_until_epochinteger | null
    The check counts while the epoch is below this; null for the first round
  • sellers[].counts_nowboolean | null
    Whether this check is the service's live verdict on-chain now (from the same live read as live; never the first round). Its SAME or DIFF is what live and the points multiplier are made of; an UNDETERMINED counts for nothing. The page's main table shows these and folds the rest below it. null when our data holds no recorded verdict for this check yet (a PENDING, or a SAME/UNDETERMINED whose status event is not read yet: see valid_until_epoch null) or the live state is missing or stale
  • sellers[].diffsinteger | null
    Answers that differed in the lead read (lead_auditor's deciding read); reads are never added
  • sellers[].totalinteger | null
    Answers compared in that read
  • sellers[].lead_auditorstring | null
    The read the row's figures come from: among the reads that agree with the verdict, the one with the largest margin (furthest over the line for a DIFF, closest to it for a SAME)
  • sellers[].auditors_splitboolean
    The auditors that got answers disagreed (one DIFF, one SAME); each read is in reads
  • sellers[].noise_ceilingnumber | null
    The lead read's 99% upper bound of the noise rate
  • sellers[].fail_lineinteger | null
    The lead read's fail line: this many different answers or more fails
  • sellers[].line_ratenumber | null
    That line as a share of the read's answers
  • sellers[].margininteger | null
    diffs − fail_line in the lead read: 0 or more fails, below 0 passes
  • sellers[].times_noisenumber | null
    The lead read's share that differed ÷ the real model's own share in it
  • sellers[].noise_diffsinteger | null
    The real model's own differences in the lead read
  • sellers[].noise_totalinteger | null
    The real model's answers compared in it
  • sellers[].reads[]array
    Each auditor's read
  • sellers[].reads[].auditorstring
    Auditor address, lowercase
  • sellers[].reads[].audited_tsinteger
    Unix seconds the report reached the chain
  • sellers[].reads[].verdictstring
    This auditor's own verdict
  • sellers[].reads[].outcomestring | null
    Why the audit could not run (the seller's error code); null when it ran
  • sellers[].reads[].first_diffsinteger | null
    Answers that differed on the first set of questions
  • sellers[].reads[].first_totalinteger | null
    Questions answered on the first set
  • sellers[].reads[].p_valuenumber | null
    Binomial p-value of the first set
  • sellers[].reads[].extra_setsinteger
    Extra sets of questions run on a close call
  • sellers[].reads[].diffsinteger | null
    Answers that differed in the deciding read (pooled over the extra sets when they were stored pooled)
  • sellers[].reads[].totalinteger | null
    Answers in the deciding read
  • sellers[].reads[].noise_diffsinteger | null
    The real model's answers that differed from each other in the same read: the noise
  • sellers[].reads[].noise_totalinteger | null
    The real model's answers compared
  • sellers[].reads[].pooled_p_valuenumber | null
    p-value of the pooled read; null without extra sets
  • sellers[].reads[].noise_ceilingnumber | null
    The 99% upper bound of the noise rate in the deciding read, what the seller is tested against
  • sellers[].reads[].proof_urlstring | null
    The seller's evidence file (questions, answers, comparison) through an IPFS gateway
  • sellers[].reads[].fail_lineinteger | null
    This many different answers or more fails the deciding read: the line the verdict used, from that read's own noise ceiling and the rule's p-value threshold
  • sellers[].reads[].margininteger | null
    diffs − fail_line: 0 or more fails, below 0 passes
  • sellers[].reads[].alphanumber | null
    The p-value the deciding read had to fall below to fail: 0.005 on a first set or a look before the last, 0.05 on the last look and in the 20-oct first round; null under a rule we do not know
  • sellers[].reads[].planned_fail_lineinteger | null
    The first set's line planned before the run, in the seller's evidence file (criticalMismatchCount); the run's own noise can move the real line above it
  • sellers[].reads[].probe_countinteger | null
    Questions in the first set
  • sellers[].reads[].powernumber | null
    Chance the first set catches a seller serving another model
  • sellers[].reads[].detailobject | null
    From the seller's own evidence file, hash-checked; null until read
  • sellers[].reads[].detail.subjects[]array
  • sellers[].reads[].detail.subjects[].subjectstring
    Subject of the questions
  • sellers[].reads[].detail.subjects[].differedinteger
    Answers that differed
  • sellers[].reads[].detail.subjects[].answeredinteger
    Answers compared
  • sellers[].reads[].detail.examples[]array
    The misses furthest from the real model
  • sellers[].reads[].detail.examples[].domainstring
    Subject
  • sellers[].reads[].detail.examples[].questionstring
    The question
  • sellers[].reads[].detail.examples[].referencestring | null
    The real model's answer (number or text)
  • sellers[].reads[].detail.examples[].answerstring | null
    The seller's answer (number or text)
  • sellers[].reads[].detail.latency_msinteger | null
    Median answer time
  • sellers[].reads[].detail.audit_cost_usdnumber | null
    What the audit spent buying from the seller
  • sellers[].reads[].detail.reasoningobject | null
    How the seller handled being asked not to reason (the reference answers without)
  • sellers[].reads[].detail.reasoning.strategystring | null
    How the auditor finally asked
  • sellers[].reads[].detail.reasoning.attempts[]array
  • sellers[].reads[].detail.reasoning.attempts[].strategystring
    A way to switch reasoning off
  • sellers[].reads[].detail.reasoning.attempts[].reasoning_tokensinteger | null
    Reasoning tokens the seller still used
  • sellers[].reads[].detail.reasoning.attempts[].valid_answersinteger | null
    Answers it could parse
  • sellers[].referencesarray | null
    What the auditors compared against
  • sellers[].references[].modelstring | null
    Reference model id
  • sellers[].references[].typestring | null
    Where it was called (openrouter)
  • sellers[].references[].provider_pin[]array
  • sellers[].references[].overridesobject | null
    Request settings the auditor forced (reasoning effort)
  • sellers[].addressstring
    Seller, lowercase 0x
  • sellers[].agent_idstring
    Agent id
  • sellers[].namestring | null
    Announced name (untrusted)
  • rounds[]array
    Newest first
  • rounds[].epochinteger
    Epoch
  • rounds[].started_tsinteger
    Unix seconds of the round's first report
  • rounds[].first_roundboolean
    Before checks counted for points
  • rounds[].passedinteger
    SAME
  • rounds[].failedinteger
    DIFF
  • rounds[].not_testedinteger
    UNDETERMINED
  • rounds[].pendinginteger
    DIFF with one report, waiting for a second

Cache 90 s. Sources: checks = chain, evidence = chain.

GET/models/videoVideo: spend, volume and models

What buyers paid for video on Antseed in the last 7 days (rolling; the home's By modality row), how much was generated, what is on sale now, the history by UTC day and every video model. Usage comes from settlements whose buyers share the per-model split: a lower bound. A settlement counts once: video when it carries seconds, else images when it carries images. The daily history starts on the first day with video; a day inside it with nothing settled is a real 0. The model list counts Organic sellers.

curl "https://antseedstats.com/api/v1/models/video"
/api/v1
Response fields (30)
  • modalitystring
    video, images or speech
  • last_7dobject
    Rolling 7 days, every seller, lower bound; a figure the cached window lacks is null
  • last_7d.spent_microinteger | null
    spent in micro-USDC (integer, 6 decimals)
  • last_7d.spent_usdstring | null
    spent in USDC, formatted with 2 decimals
  • last_7d.videosinteger | null
    Videos delivered
  • last_7d.video_secondsinteger | null
    Seconds of video
  • on_saleobject
    From our DHT crawl
  • on_sale.modelsinteger
    Models on sale now (name variants count once)
  • on_sale.sellersinteger
    Distinct sellers advertising them
  • days[]array
    From the first day with this modality to today (partial), oldest first
  • days[].daystring
    UTC day, YYYY-MM-DD
  • days[].spent_microinteger | null
    spent in micro-USDC (integer, 6 decimals)
  • days[].spent_usdstring | null
    spent in USDC, formatted with 2 decimals
  • days[].videosinteger | null
    Videos delivered
  • days[].video_secondsinteger | null
    Seconds of video
  • models[]array
    Organic sales of the last 7 days first (null when that window could not be read), then lifetime sales
  • models[].modelstring
    Canonical id
  • models[].displaystring
    Name
  • models[].servedboolean
    A seller announces it now; false = usage history only
  • models[].sellersinteger
    Sellers advertising it
  • models[].min_video_usd_per_snumber | null
    Cheapest price per second of video, 720p
  • models[].min_video_usdnumber | null
    Cheapest price per video
  • models[].settled_7d_microinteger | null
    settled 7d in micro-USDC (integer, 6 decimals)
  • models[].settled_7d_usdstring | null
    settled 7d in USDC, formatted with 2 decimals
  • models[].usageobject
    Lifetime, Organic sellers, lower bound
  • models[].usage.settled_microinteger
    settled in micro-USDC (integer, 6 decimals)
  • models[].usage.settled_usdstring
    settled in USDC, formatted with 2 decimals
  • models[].usage.buyersinteger
    Buyers
  • models[].usage.video_secondsinteger
    Seconds of video
  • models[].usage.videosinteger
    Videos delivered

Cache 5 to 8 min. Sources: usage = chain, on_sale = announce, prices = announce, organic = antseedstats.

GET/models/imagesImages: spend, volume and models

What buyers paid for images on Antseed in the last 7 days (rolling; the home's By modality row), how much was generated, what is on sale now, the history by UTC day and every image model. Usage comes from settlements whose buyers share the per-model split: a lower bound. A settlement counts once: video when it carries seconds, else images when it carries images. The daily history starts on the first day with images; a day inside it with nothing settled is a real 0. The model list counts Organic sellers.

curl "https://antseedstats.com/api/v1/models/images"
/api/v1
Response fields (27)
  • modalitystring
    video, images or speech
  • last_7dobject
    Rolling 7 days, every seller, lower bound; a figure the cached window lacks is null
  • last_7d.spent_microinteger | null
    spent in micro-USDC (integer, 6 decimals)
  • last_7d.spent_usdstring | null
    spent in USDC, formatted with 2 decimals
  • last_7d.imagesinteger | null
    Images generated
  • on_saleobject
    From our DHT crawl
  • on_sale.modelsinteger
    Models on sale now (name variants count once)
  • on_sale.sellersinteger
    Distinct sellers advertising them
  • days[]array
    From the first day with this modality to today (partial), oldest first
  • days[].daystring
    UTC day, YYYY-MM-DD
  • days[].spent_microinteger | null
    spent in micro-USDC (integer, 6 decimals)
  • days[].spent_usdstring | null
    spent in USDC, formatted with 2 decimals
  • days[].imagesinteger | null
    Images generated
  • models[]array
    Organic sales of the last 7 days first (null when that window could not be read), then lifetime sales
  • models[].modelstring
    Canonical id
  • models[].displaystring
    Name
  • models[].servedboolean
    A seller announces it now; false = usage history only
  • models[].sellersinteger
    Sellers advertising it
  • models[].min_image_usdnumber | null
    Cheapest price per image
  • models[].save_pctnumber | null
    Cheapest image price below the Venice list price, percent (100 when free); null without a reference or a saving
  • models[].settled_7d_microinteger | null
    settled 7d in micro-USDC (integer, 6 decimals)
  • models[].settled_7d_usdstring | null
    settled 7d in USDC, formatted with 2 decimals
  • models[].usageobject
    Lifetime, Organic sellers, lower bound
  • models[].usage.settled_microinteger
    settled in micro-USDC (integer, 6 decimals)
  • models[].usage.settled_usdstring
    settled in USDC, formatted with 2 decimals
  • models[].usage.buyersinteger
    Buyers
  • models[].usage.imagesinteger
    Images

Cache 5 to 8 min. Sources: usage = chain, on_sale = announce, prices = announce, organic = antseedstats.

GET/models/speechSpeech: spend, volume and models

What buyers paid for speech on Antseed in the last 7 days (rolling; the home's By modality row), how much was generated, what is on sale now, the history by UTC day and every speech model. Usage comes from settlements whose buyers share the per-model split: a lower bound. A settlement counts once: speech when it carries characters and no images or video (metadata v5, since 4 November 2026). The daily history starts on the first day with speech; a day inside it with nothing settled is a real 0. The model list counts Organic sellers.

curl "https://antseedstats.com/api/v1/models/speech"
/api/v1
Response fields (26)
  • modalitystring
    video, images or speech
  • last_7dobject
    Rolling 7 days, every seller, lower bound; a figure the cached window lacks is null
  • last_7d.spent_microinteger | null
    spent in micro-USDC (integer, 6 decimals)
  • last_7d.spent_usdstring | null
    spent in USDC, formatted with 2 decimals
  • last_7d.charactersinteger | null
    Characters read aloud
  • on_saleobject
    From our DHT crawl
  • on_sale.modelsinteger
    Models on sale now (name variants count once)
  • on_sale.sellersinteger
    Distinct sellers advertising them
  • days[]array
    From the first day with this modality to today (partial), oldest first
  • days[].daystring
    UTC day, YYYY-MM-DD
  • days[].spent_microinteger | null
    spent in micro-USDC (integer, 6 decimals)
  • days[].spent_usdstring | null
    spent in USDC, formatted with 2 decimals
  • days[].charactersinteger | null
    Characters read aloud
  • models[]array
    Organic sales of the last 7 days first (null when that window could not be read), then lifetime sales
  • models[].modelstring
    Canonical id
  • models[].displaystring
    Name
  • models[].servedboolean
    A seller announces it now; false = usage history only
  • models[].sellersinteger
    Sellers advertising it
  • models[].min_usd_per_1m_charsnumber | null
    Cheapest price per 1M characters read aloud
  • models[].save_pctnumber | null
    Cheapest character price below the Venice list price per 1M characters, percent (100 when free); null without a reference or a saving
  • models[].settled_7d_microinteger | null
    settled 7d in micro-USDC (integer, 6 decimals)
  • models[].settled_7d_usdstring | null
    settled 7d in USDC, formatted with 2 decimals
  • models[].usageobject
    Lifetime, Organic sellers, lower bound
  • models[].usage.settled_microinteger
    settled in micro-USDC (integer, 6 decimals)
  • models[].usage.settled_usdstring
    settled in USDC, formatted with 2 decimals
  • models[].usage.buyersinteger
    Buyers

Cache 5 to 8 min. Sources: usage = chain, on_sale = announce, prices = announce, organic = antseedstats.

GET/models/{id}Where to buy one model

Accepts the canonical id or any advertised spelling, in any case or separator (Kimi K3 works). Sellers are deduplicated (cheapest input price) and come trusted first, then cheapest, unless sort says otherwise. The filters answer the question a buyer asks: which sellers of this model meet my conditions. Without filters the list hides nobody.

curl "https://antseedstats.com/api/v1/models/kimi-k3?trusted=true"
/api/v1
Parameters
  • idpath · string · required
    Canonical model id or an advertised name
  • freequery · bool
    true: only free listings; false: only paid
  • min_reputationquery · number
    Only sellers with a trust score at or above this (0 to 100)
  • verifiedquery · bool
    true: only sellers with a domain or GitHub proof we verified
  • organicquery · bool
    true: only sellers outside the wash registry and our confirmed clusters; false: only the others
  • livequery · bool
    true: only sellers announcing on the DHT right now
  • trustedquery · bool
    true: only sellers that settled on-chain and score at or above 60, the SDK's own routing gate (the calculator's rule)
  • max_input_usd_per_mquery · number
    Only listings at or below this USD per 1M input tokens
  • max_output_usd_per_mquery · number
    Only listings at or below this USD per 1M output tokens
  • sortquery · trusted | price | reputation · default trusted
    trusted: trusted sellers first, then cheapest input; price: cheapest input first; reputation: highest trust score first
  • seller_daysquery · bool
    true: each seller_share entry adds days, what that seller sold each day (the page's Daily usage by seller; about doubles the response)
Response fields (126)
  • modelstring
    Canonical id
  • displaystring
    Name
  • customboolean
    Seller's own model
  • advertised_as[]array
  • usageobject
    Lifetime, every seller, lower bound
  • usage.settled_microinteger
    settled in micro-USDC (integer, 6 decimals)
  • usage.settled_usdstring
    settled in USDC, formatted with 2 decimals
  • usage.tokensinteger
    Tokens
  • usage.imagesinteger
    Images
  • usage.video_secondsinteger
    Seconds of video
  • usage.videosinteger
    Videos delivered
  • usage.requestsinteger
    Requests
  • usage.buyersinteger
    Buyers
  • usage.sellersinteger
    Sellers
  • usage_organic.settled_microinteger
    settled in micro-USDC (integer, 6 decimals)
  • usage_organic.settled_usdstring
    settled in USDC, formatted with 2 decimals
  • usage_organic.tokensinteger
    Tokens
  • usage_organic.imagesinteger
    Images
  • usage_organic.video_secondsinteger
    Seconds of video
  • usage_organic.videosinteger
    Videos delivered
  • usage_organic.requestsinteger
    Requests
  • usage_organic.buyersinteger
    Buyers
  • usage_organic.sellersinteger
    Sellers
  • days_organicarray | null
    This model's usage by day, Organic sellers only (the page's Daily usage card). Buyer opt-in: a lower bound. null when it could not be read
  • days_organic[].daystring
    UTC day, YYYY-MM-DD; continuous from the first day with usage, a day without reports is a real 0
  • days_organic[].settled_microinteger
    settled in micro-USDC (integer, 6 decimals)
  • days_organic[].settled_usdstring
    settled in USDC, formatted with 2 decimals
  • days_organic[].tokensinteger
    Paid tokens, Organic sellers
  • days_organic[].free_tokensinteger
    Free-tier tokens, Organic sellers
  • days_organic[].imagesinteger
    Images
  • days_organic[].video_secondsinteger
    Seconds of video
  • days_organic[].videosinteger
    Videos delivered
  • days_organic[].buyersinteger
    Wallets that paid an Organic seller
  • days_organic[].free_only_buyersinteger
    Wallets that only used an Organic seller for free
  • days_organic[].sellersinteger
    Organic sellers paid that day
  • days_organic[].free_only_sellersinteger
    Organic sellers that only served free that day
  • seller_sharearray | null
    What every Organic seller sold of this model, listed now or not, biggest earner first (the page's Top sellers donut). Buyer opt-in: a lower bound. null when it could not be read
  • seller_share[].addressstring
    Lowercase 0x
  • seller_share[].namestring | null
    Announced name (untrusted)
  • seller_share[].allobject
    Lifetime
  • seller_share[].all.settled_microinteger
    settled in micro-USDC (integer, 6 decimals)
  • seller_share[].all.settled_usdstring
    settled in USDC, formatted with 2 decimals
  • seller_share[].all.tokensinteger
    Paid tokens
  • seller_share[].all.free_tokensinteger
    Free-tier tokens
  • seller_share[].all.imagesinteger
    Images
  • seller_share[].all.video_secondsinteger
    Seconds of video
  • seller_share[].all.videosinteger
    Videos delivered
  • seller_share[].last_7dobject | null
    The last 7 UTC days, today included; null on a value cached before these fields existed
  • seller_share[].last_7d.settled_microinteger
    settled in micro-USDC (integer, 6 decimals)
  • seller_share[].last_7d.settled_usdstring
    settled in USDC, formatted with 2 decimals
  • seller_share[].last_7d.tokensinteger
    Paid tokens
  • seller_share[].last_7d.free_tokensinteger
    Free-tier tokens
  • seller_share[].last_7d.imagesinteger
    Images
  • seller_share[].last_7d.video_secondsinteger
    Seconds of video
  • seller_share[].last_7d.videosinteger
    Videos delivered
  • seller_share[].last_30dobject | null
    The last 30 UTC days, today included
  • seller_share[].last_30d.settled_microinteger
    settled in micro-USDC (integer, 6 decimals)
  • seller_share[].last_30d.settled_usdstring
    settled in USDC, formatted with 2 decimals
  • seller_share[].last_30d.tokensinteger
    Paid tokens
  • seller_share[].last_30d.free_tokensinteger
    Free-tier tokens
  • seller_share[].last_30d.imagesinteger
    Images
  • seller_share[].last_30d.video_secondsinteger
    Seconds of video
  • seller_share[].last_30d.videosinteger
    Videos delivered
  • seller_share[].last_90dobject | null
    The last 90 UTC days, today included
  • seller_share[].last_90d.settled_microinteger
    settled in micro-USDC (integer, 6 decimals)
  • seller_share[].last_90d.settled_usdstring
    settled in USDC, formatted with 2 decimals
  • seller_share[].last_90d.tokensinteger
    Paid tokens
  • seller_share[].last_90d.free_tokensinteger
    Free-tier tokens
  • seller_share[].last_90d.imagesinteger
    Images
  • seller_share[].last_90d.video_secondsinteger
    Seconds of video
  • seller_share[].last_90d.videosinteger
    Videos delivered
  • seller_share[].daysarray | null
    Each UTC day this seller sold or served the model, oldest first, days with nothing left out; values to 4 significant digits (the page's Daily usage by seller). Only with seller_days=true; null on a value cached before this field existed
  • seller_share[].days[].daystring
    UTC day, YYYY-MM-DD
  • seller_share[].days[].settled_microinteger
    settled in micro-USDC (integer, 6 decimals)
  • seller_share[].days[].settled_usdstring
    settled in USDC, formatted with 2 decimals
  • seller_share[].days[].tokensinteger
    Paid tokens
  • seller_share[].days[].free_tokensinteger
    Free-tier tokens
  • seller_share[].days[].imagesinteger
    Images
  • seller_share[].days[].video_secondsinteger
    Seconds of video
  • sellers[]array
  • sellers[].addressstring | null
    Lowercase 0x
  • sellers[].agent_idstring
    Agent id
  • sellers[].namestring | null
    Announced name (untrusted)
  • sellers[].advertised_asstring
    Raw name
  • sellers[].input_usd_per_mnumber | null
    USD per 1M input
  • sellers[].cached_input_usd_per_mnumber | null
    USD per 1M cached input
  • sellers[].output_usd_per_mnumber | null
    USD per 1M output
  • sellers[].image_usdnumber | null
    USD per image
  • sellers[].video_usd_per_snumber | null
    USD per second of video, 720p
  • sellers[].video_usdnumber | null
    USD per video
  • sellers[].usd_per_1m_charsnumber | null
    Speech: USD per 1M input characters (the text read aloud); null when not a speech listing
  • sellers[].speech_voicesarray | null
    Speech: the voices it accepts; another voice is refused unless speech_custom_voices
  • sellers[].speech_formatsarray | null
    Speech: the response formats it returns
  • sellers[].speech_custom_voicesboolean | null
    Speech: also takes account-specific voice ids
  • sellers[].model_checkstring | null
    Antseed's live model check on this seller's listing of the model: SAME or DIFF; null when none is live
  • sellers[].freeboolean
    0 and 0; speech: the character price is 0; never for video
  • sellers[].save_pctnumber | null
    Percent below the model's list price, as the page's Save column (1-99, 100 when free). Text: input rate vs the OpenRouter list input rate, or Venice's when OpenRouter has none. Image models: image_usd vs the Venice list price per image. Speech models: usd_per_1m_chars vs the Venice list price per 1M characters. Null when not cheaper, no list price, or a video model
  • sellers[].reputationnumber | null
    Trust score
  • sellers[].verifiedboolean
    Verified proof
  • sellers[].organicboolean
    In our Organic view
  • sellers[].provenboolean
    Has settled on-chain (carries a trust score and usage)
  • sellers[].trustedboolean
    Proven and scoring at or above 60, the SDK routing gate (our rule, the calculator's pick)
  • sellers[].liveboolean
    Announcing on the DHT right now
  • sellers[].request_idstring | null
    peerId@model, what a buyer sends
  • sellers[].healthstring | null
    Seller-reported state of this service: healthy, failing or unknown. null when the seller reports none (SDK older than 0.2.157), which is not the same as healthy
  • sellers[].health_reasonstring | null
    Why it fails, the SDK's code: timeout, rate_limited, out_of_credit, auth, model_unavailable, upstream_error, unreachable, init_failed
  • sellers[].health_sourcestring | null
    probe (the seller's own 5-minute check) or traffic (live buyer requests)
  • sellers[].health_checked_tsinteger | null
    Unix seconds of the evidence behind health
  • sellers[].health_since_tsinteger | null
    Unix seconds of the first failure of the current streak
  • sellers[].health_http_statusinteger | null
    Upstream HTTP status of the last failed check
  • sellers[].health_not_servedboolean
    Failing for a day or more: the seller stopped serving it; out of every count and catalog, listed for the record
  • sellers[].requests_last_hourinteger | null
    Buyer requests in the seller's last hour, as the seller counts them
  • sellers[].success_ratenumber | null
    Share of them answered without a server-side failure, 0 to 1
  • sellers[].first_token_ms_p50integer | null
    Median ms to the first streamed token, seller-measured
  • sellers[].first_token_ms_p90integer | null
    90th percentile ms to the first token
  • sellers[].tokens_per_second_p50integer | null
    Median output speed, tokens per second
  • sellers[].usageobject | null
    What this seller sold of the model, lifetime, lower bound; null when nothing is attributed or the seller is outside the Organic view
  • sellers[].usage.settled_microinteger
    settled in micro-USDC (integer, 6 decimals)
  • sellers[].usage.settled_usdstring
    settled in USDC, formatted with 2 decimals
  • sellers[].usage.paidboolean
    Has a paid report for this model
  • sellers[].usage.freeboolean
    Has a free report for this model
  • sellers[].usage.tokensinteger
    Paid tokens
  • sellers[].usage.free_tokensinteger
    Free-tier tokens
  • sellers[].usage.imagesinteger
    Images
  • sellers[].usage.video_secondsinteger
    Seconds of video
  • sellers[].usage.videosinteger
    Videos delivered

Cache 90 s. Sources: prices = announce, reputation = antseedstats, organic = antseedstats, usage = chain.

GET/pricesEvery listing with list-price savings

The network explorer flattened: one row per seller and model with the advertised prices, whether it is free, the OpenRouter list price for the same model and the saving against it. Filter by model to compare sellers; the intent filters (trusted, verified, organic, live, price caps) narrow to the sellers a buyer would use. Sorted by model, then trusted first and cheapest input.

curl "https://antseedstats.com/api/v1/prices?model=kimi-k3&trusted=true"
/api/v1
Parameters
  • modelquery · string
    Canonical model id to filter on (any spelling)
  • freequery · bool
    true: only free listings; false: only paid
  • viewquery · all | organic · default all
    organic leaves out sellers in the on-chain wash registry plus our confirmed clusters; all keeps every seller
  • min_reputationquery · number
    Only sellers with a trust score at or above this (0 to 100)
  • verifiedquery · bool
    true: only sellers with a domain or GitHub proof we verified
  • organicquery · bool
    true: only sellers outside the wash registry and our confirmed clusters; false: only the others
  • livequery · bool
    true: only sellers announcing on the DHT right now
  • trustedquery · bool
    true: only sellers that settled on-chain and score at or above 60, the SDK's own routing gate (the calculator's rule)
  • max_input_usd_per_mquery · number
    Only listings at or below this USD per 1M input tokens
  • max_output_usd_per_mquery · number
    Only listings at or below this USD per 1M output tokens
  • sortquery · trusted | price | reputation · default trusted
    trusted: trusted sellers first, then cheapest input; price: cheapest input first; reputation: highest trust score first
  • limitquery · int · default 100
    Rows to return (1 to 1000)
  • offsetquery · int · default 0
    Rows to skip
Response fields (26)
  • totalinteger
    Listings matching
  • listings[]array
  • listings[].sellerstring
    Lowercase 0x address
  • listings[].seller_namestring | null
    Announced display name (untrusted text)
  • listings[].agent_idstring | null
    On-chain agent id
  • listings[].modelstring
    Canonical model id (the /models/{id} slug)
  • listings[].model_displaystring
    Display name
  • listings[].advertised_asstring
    The raw name the seller announces
  • listings[].input_usd_per_mnumber | null
    USD per 1M input tokens
  • listings[].cached_input_usd_per_mnumber | null
    USD per 1M cached input tokens
  • listings[].output_usd_per_mnumber | null
    USD per 1M output tokens
  • listings[].freeboolean
    Input and output both priced at 0
  • listings[].list_input_usd_per_mnumber | null
    List price for the same model, USD per 1M input: OpenRouter, or Venice when OpenRouter has none
  • listings[].list_sourcestring | null
    Where the list price comes from
  • listings[].save_pctinteger | null
    Percent below the list price, 1 to 99, only when the listing beats it
  • listings[].reputationnumber | null
    Trust score 0 to 100
  • listings[].verifiedboolean
    Has a domain or GitHub proof we verified
  • listings[].organicboolean
    Seller in our Organic view
  • listings[].provenboolean
    Has settled on-chain (carries a trust score and usage)
  • listings[].trustedboolean
    Proven and scoring at or above 60, the SDK routing gate (our rule, the calculator's pick)
  • listings[].liveboolean
    Announcing on the DHT right now
  • listings[].stake_usdcnumber | null
    Seller's USDC stake on-chain (proven peers)
  • listings[].capabilities[]array
  • listings[].healthstring | null
    Seller-reported state of this service: healthy, failing or unknown. null when the seller reports none, which is not the same as healthy
  • listings[].health_reasonstring | null
    Why it fails, the SDK's code (out_of_credit, rate_limited, auth, model_unavailable, upstream_error, timeout, unreachable, init_failed)
  • listings[].health_since_tsinteger | null
    Unix seconds of the first failure of the current streak

Cache 90 s. Sources: prices = announce, list_price = antseedstats, reputation = antseedstats, organic = antseedstats.

GET/labsThe labs behind the models

One row per model lab (organisation) with models on Antseed: the name, domain and logo come from the model catalogue we enrich with, and models counts canonical models of that lab advertised by at least one seller. Sorted by model count.

curl "https://antseedstats.com/api/v1/labs"
/api/v1
Response fields (7)
  • totalinteger
    Labs
  • labs[]array
  • labs[].orgstring
    Lab key (the /labs/{org} slug, lowercase)
  • labs[].namestring
    Display name
  • labs[].domainstring | null
    Official domain
  • labs[].logostring | null
    Logo image URL
  • labs[].modelsinteger
    Canonical models advertised

Cache 90 s. Sources: labs = antseedstats, models = announce.

GET/labs/{org}One lab: its models on Antseed

The lab's profile from the model catalogue (bio, links) and every canonical model of that lab with the number of sellers advertising it and the cheapest advertised prices.

curl "https://antseedstats.com/api/v1/labs/moonshot"
/api/v1
Parameters
  • orgpath · string · required
    Lab key, the slug of /labs/{org} (moonshot, deepseek, alibaba, ...; lowercase, may contain a space)
Response fields (11)
  • orgstring
    Lab key
  • namestring
    Display name
  • domainstring | null
    Official domain
  • biostring | null
    Short description
  • logostring | null
    Logo image URL
  • linksobject
    Named links (official, github, huggingface, x, ...)
  • models[]array
  • models[].modelstring
    Canonical id
  • models[].sellersinteger
    Sellers advertising it
  • models[].min_input_usd_per_mnumber | null
    Cheapest input price
  • models[].min_output_usd_per_mnumber | null
    Cheapest output price

Cache 90 s. Sources: lab = antseedstats, models = announce.

Staking and rewards
GET/staking/poolsEvery provider pool: stake, power, reward

The $ANTS staking market as read from the pools contracts: one row per provider pool (agent id) with its stake and power this epoch, the same for the next epoch (staking has a one-epoch activation delay), its usage points and staker reward in the last complete epoch, its recognized volume over the last complete epochs and so far in this one, and what a new 10,000 ANTS stake would earn there: per week at a 1-week lock and at the maximum lock, as a simple (not compounded) APR, and projected for a stake made now. What one unit of power earns depends only on the pool's volume, not on how much is already staked in it. Figures tie out against the contract totals; degraded says when they did not.

curl "https://antseedstats.com/api/v1/staking/pools"
/api/v1
Parameters
  • viewquery · all | organic · default all
    organic leaves out sellers in the on-chain wash registry plus our confirmed clusters; all keeps every seller
Response fields (96)
  • epochinteger
    Current epoch (stake and power columns)
  • next_epochinteger
    Next epoch (next_* columns)
  • reward_epochinteger
    Last complete epoch (usage and reward columns)
  • degradedboolean
    True when the market did not tie out and the figures are not shown
  • reasonstring | null
    Why, when degraded
  • total_stake_antsnumber | null
    Active stake, all pools; null when degraded
  • total_principal_antsnumber | null
    Principal, all pools
  • total_power_antsnumber | null
    Power, all pools
  • next_total_stake_antsnumber | null
    Stake from the next epoch
  • next_total_power_antsnumber | null
    Power from the next epoch
  • restake_bonus_pctnumber | null
    Weight bonus the contract grants restaked rewards, %
  • max_lock_epochsinteger | null
    Longest lock
  • budget_antsnumber | null
    Staker budget of the reward epoch
  • reward_per_power_antsnumber | null
    Budget over the reward epoch's power
  • claimable_antsnumber | null
    Settled rewards not yet withdrawn, all pools
  • unsettled_antsnumber | null
    Earned in the reward epoch and never settled on-chain
  • reward_flows_as_of_blockstring | null
    Block every row's reward_flows describes (the lower of the flows and stake-lineage indexers' cursors); null when one has not run
  • reward_flows_degradedstring | null
    Why every reward_flows is null, when they are: not-indexed, tie-out (the network identity failed at that block), balance-unread, lineage-lagging (the stake-lineage indexer trails), lineage-gap (a restaked reward ends in a position closed at that block with no recorded exit), read-failed
  • poolsinteger | null
    Pools listed; null when degraded
  • positionsinteger | null
    Active positions
  • stakersinteger | null
    Distinct active stakers
  • returns_stake_antsnumber | null
    The new stake the weekly and APR columns describe (fresh, no restake bonus)
  • returns_lock_epochsobject | null
    The two lock ends used
  • returns_lock_epochs.one_weekinteger | null
    Epochs of the 1-week lock; null when the contract would not accept it
  • returns_lock_epochs.max_lockinteger
    Epochs of the maximum lock
  • projectionobject | null
    Scenario behind the projected_* columns: that epoch's budget and every pool's power as already booked for it, with basis's volume mix taken as a whole epoch. It assumes the mix holds and nothing more is staked. Null when an input failed to read or tie out
  • projection.epochinteger
    The epoch a stake made now counts from
  • projection.basisstring
    Volume mix used: live (the current epoch, from 48 h in) or last (reward_epoch)
  • projection.budget_antsnumber
    That epoch's staker budget
  • rows[]array
  • rows[].agent_idinteger
    On-chain agent id of the pool
  • rows[].sellerstring | null
    Seller address the directory maps the agent to
  • rows[].namestring | null
    Announced display name (untrusted)
  • rows[].organicboolean
    Seller outside the wash registry and our confirmed clusters (our judgement)
  • rows[].stake_antsnumber
    Active stake this epoch (principal plus restake bonus)
  • rows[].principal_antsnumber
    What stakers put in
  • rows[].power_antsnumber
    Stake times epochs remaining
  • rows[].power_share_pctnumber
    Share of all pools' power, %
  • rows[].next_stake_antsnumber
    Stake from the next epoch
  • rows[].next_principal_antsnumber
    Principal from the next epoch
  • rows[].next_power_antsnumber
    Power from the next epoch
  • rows[].next_power_share_pctnumber
    Share of next epoch's power, %
  • rows[].usage_pointsnumber
    Raw pool points in the last complete epoch
  • rows[].usage_share_pctnumber
    Share of all pools' points, %
  • rows[].reward_antsnumber
    Last complete epoch's staker reward for this pool
  • rows[].reward_settledboolean
    Settled by the contract (false: computed from its inputs)
  • rows[].claimable_antsnumber
    Settled and not yet withdrawn
  • rows[].unused_antsnumber | null
    Earned rewards not yet turned into stake
  • rows[].unused_power_antsnumber | null
    What unused would be worth as power at the maximum lock
  • rows[].reward_claimed_pctnumber | null
    Of reward_ants, the share its positions have already taken (restaked or withdrawn), weighted by the power each held that epoch; null until settled
  • rows[].reward_unclaimed_antsnumber | null
    reward_ants still sitting for its stakers to take; null until settled
  • rows[].stays_next_pctnumber | null
    Share of the active principal that still counts next epoch
  • rows[].top_staker_pctnumber | null
    The largest staker's share of the active principal
  • rows[].median_lock_epochsinteger | null
    Principal-weighted median of the epochs left on the locks
  • rows[].arrived_this_epoch_pctnumber | null
    Share of the active principal that activated this epoch
  • rows[].restaked_pctnumber | null
    Share of the active principal funded by pool rewards (restaked)
  • rows[].reward_flowsobject | null
    Null when the flows indexer has not run, a settled cell could not be read, the network identity (minted into the rewards contract minus every exit equals its balance, at the same block, to the wei) failed, or a restaked reward could not be followed to an open position; see reward_flows_as_of_block and reward_flows_degraded at the top
  • rows[].reward_flows.earned_to_date_antsnumber
    Σ settled staker rewards of this pool, first staker epoch to date
  • rows[].reward_flows.restaked_here_antsnumber
    Restaked by restakeStakerRewards and still staked in this pool now, followed through moves, merges and splits
  • rows[].reward_flows.restaked_elsewhere_antsnumber
    Restaked, then moved to another pool (moveStake needs no claim; the contract's move penalty is 0)
  • rows[].reward_flows.withdrawn_antsnumber
    Claimed to a wallet by claimStakerRewards, plus restaked and later withdrawn from staking
  • rows[].reward_flows.restaked_then_withdrawn_antsnumber | null
    The part of withdrawn_ants that was restaked first and then withdrawn from staking (withdrawStake), slashing included; null on a snapshot older than this field
  • rows[].reward_flows.waiting_antsnumber
    Settled and not yet taken (earned minus the three above)
  • rows[].reward_flows.restaked_here_pctnumber | null
    Share of earned_to_date_ants; null when nothing was earned
  • rows[].reward_flows.restaked_elsewhere_pctnumber | null
    Share of earned_to_date_ants
  • rows[].reward_flows.withdrawn_pctnumber | null
    Share of earned_to_date_ants
  • rows[].reward_flows.waiting_pctnumber | null
    Share of earned_to_date_ants
  • rows[].reward_flows.received_from_elsewhere_antsnumber
    Rewards earned and restaked in other pools that sit in this pool now, moved here (not part of the shares)
  • rows[].reward_flows.claimsinteger
    claimStakerRewards exits counted
  • rows[].reward_flows.restakesinteger
    restakeStakerRewards exits counted
  • rows[].volume_usdc_by_epoch[]array
    Recognized volume (agentPoolPointsByEpoch: USDC after the protocol's wash and verification policies) of the last complete epochs, oldest first, ending at reward_epoch, never before epoch 22
  • rows[].volume_usdc_by_epoch[].epochinteger
    Complete epoch
  • rows[].volume_usdc_by_epoch[].usdcnumber | null
    Recognized volume, USDC; null when that read failed
  • rows[].volume_change_pctnumber | null
    reward_epoch's volume against the epoch before, both complete, %; null when new or unknown
  • rows[].volume_newboolean
    No volume the epoch before, some in reward_epoch
  • rows[].live_volume_usdcnumber | null
    Recognized volume so far in the current epoch (partial); null when the read failed
  • rows[].reward_per_1k_power_antsnumber | null
    Marginal rate in reward_epoch: budget × volume × 1000 / Σ(volume × power), ANTS per 1,000 power for the week; lock-independent
  • rows[].weekly_ants_10k_one_weeknumber | null
    What a new fresh 10,000 ANTS stake locked one week would have earned in reward_epoch, exact formula budget × U × p / (S + U × p), ANTS
  • rows[].weekly_ants_10k_max_locknumber | null
    The same at the maximum lock (power = stake × max_lock_epochs), ANTS
  • rows[].apr_one_week_pctnumber | null
    weekly_ants_10k_one_week as a simple yearly rate (× 365/7, not compounded), %
  • rows[].apr_max_lock_pctnumber | null
    The same at the maximum lock, %
  • rows[].weekly_ants_10k_max_lock_nownumber | null
    weekly_ants_10k_max_lock at the seller's current model-check multiplier, reward_epoch's volume held: × multiplier × Σ points / Σ (multiplier × points) over all pools. The page shows and ranks on it; a pool at ×0 is 0. A live check, not a ban: it follows the live verdicts (two epochs each)
  • rows[].apr_max_lock_pct_nownumber | null
    apr_max_lock_pct on the same basis, %
  • rows[].model_checksobject | null
    Antseed's model checks, read live off AntseedVerificationResults; null when the seller has no live verdict or the state could not be read (see the page: never shown as x0 on a failed read)
  • rows[].model_checks.sameinteger
    Live SAME verdicts (answers as the advertised model), every service of the agent
  • rows[].model_checks.diffinteger
    Live DIFF verdicts (very likely another model; two auditors agreed)
  • rows[].model_checks.points_multipliernumber | null
    The verification policy's multiplier on this seller's, its buyers' and its pool's points from each settlement on (rates read from the policy); null when no policy is registered
  • rows[].model_checks.read_tsinteger
    UTC seconds of the read (every 5 min)
  • rows[].projected_weekly_ants_10k_max_locknumber | null
    weekly_ants_10k_max_lock for a stake made now (it counts from next_epoch), under projection's scenario; null when there is no projection
  • rows[].projected_apr_one_week_pctnumber | null
    Projected simple APR at the 1-week lock, %
  • rows[].projected_apr_max_lock_pctnumber | null
    Projected simple APR at the maximum lock, %
  • rows[].apy_one_week_pctnumber | null
    Deprecated (29-oct-2026, use apr_*): the old estimate, a 1-week lock compounded weekly, %
  • rows[].apy_two_years_pctnumber | null
    Deprecated: the old compounded estimate at the longest lock, %
  • rows[].positionsinteger
    Active positions
  • rows[].stakersinteger
    Distinct active stakers
  • rows[].pending_positionsinteger
    Positions waiting for the next epoch

Cache 5 to 8 min. Sources: pools = chain, volume = chain, name = announce, organic = antseedstats, returns = antseedstats, projection = antseedstats, apy = antseedstats.

GET/staking/freeFree AI with stake: every provider's offer

Antseed providers can give free AI to buyers who lock ANTS in their pool. One row per live offer, read from each provider's own signed announcement: the base free requests everyone gets (per buyer address and per IP, per window) and the stake tiers that multiply it or lift the cap, with the ANTS each tier takes on a one-year or a max lock. A row covers either the provider's free models (one shared allowance) or one model with its own program (on a paid model, a free allowance; stakers_only means nothing free below the first tier). Pass a stake (ants, lock_weeks or max_lock) and each row says what it gets: power is ANTS times weeks of lock, counts from next_epoch, only with that provider, and falls every week, so holds_weeks says how long the tier lasts (null with a max lock). Providers can change their terms at any time.

curl "https://antseedstats.com/api/v1/staking/free?ants=10000&lock_weeks=52"
/api/v1
Parameters
  • antsquery · number · default 10000
    ANTS to stake (a fresh stake, no restake bonus)
  • lock_weeksquery · int · default 52
    Weeks of lock (one epoch is a week); ignored with max_lock
  • max_lockquery · bool
    true: a max lock, which holds 104 weeks of power while it stays on
  • sellerquery · address
    Only this provider's offers
  • modelquery · string
    Only offers covering a model whose id contains this text
  • addressquery · address
    A staker: each row adds your_stake, what the positions this address holds in that provider's pool get, as the provider counts them (its own pool only, from the epoch a position activates, the strongest 64). Stake another wallet shares by a signed code is off-chain and not counted.
Response fields (55)
  • next_epochinteger
    The epoch a stake made now starts counting
  • antsnumber
    The stake described
  • lock_weeksinteger
    Its lock
  • max_lockboolean
    Max lock
  • powernumber
    ANTS × weeks of lock, in ANTS-epochs (the unit of min_power)
  • addressstring | null
    The staker asked for (lowercase); null without address
  • epochinteger
    The current epoch, which your_stake describes
  • stake_readboolean | null
    The pools were read for address; false means its stake is unknown, not empty; null without address
  • staked_antsnumber | null
    What address has staked in every pool, offers or not; null without address or unread
  • staked_poolsinteger | null
    Pools it has stake in; null without address or unread
  • providersinteger
    Providers listed
  • offersinteger
    Offers listed
  • rows[]array
  • rows[].your_stakeobject | null
    What the stake address holds in this provider's pool gets; null without address
  • rows[].your_stake.antsnumber
    Principal address holds in this provider's pool
  • rows[].your_stake.positionsinteger
    Positions counted
  • rows[].your_stake.powernumber
    Their power this epoch, ANTS-epochs
  • rows[].your_stake.tierinteger | null
    Tier reached now (1-based); null = base or none
  • rows[].your_stake.freeboolean
    Gets free requests now
  • rows[].your_stake.requestsinteger | null
    Free requests per window now; null = unlimited
  • rows[].your_stake.unlimitedboolean
    No cap now
  • rows[].your_stake.next_epoch_requestsinteger | null
    The same from the next epoch (a position made this epoch counts then); null = unlimited
  • rows[].your_stake.next_epoch_unlimitedboolean
    No cap from the next epoch
  • rows[].your_stake.drops_epochinteger | null
    First epoch the tier drops as the lock runs down; null while it holds or there is none
  • rows[].your_stake.requests_afterinteger | null
    Free requests per window from drops_epoch; null = unlimited or no drop
  • rows[].your_stake.next_tier_antsinteger | null
    ANTS to add in this pool, max-locked, to reach the next tier; null at the top
  • rows[].sellerstring
    Lowercase 0x
  • rows[].namestring | null
    Announced name (untrusted)
  • rows[].agent_idinteger | null
    Pool (agent id); null when not in the directory
  • rows[].reputationnumber | null
    The provider's buyer trust score, 0-100 (the SDK's formula, computed by us); null when not scored
  • rows[].stake_urlstring | null
    Antseed's own stake form opened on this pool (antseedpoints.com); null without a pool
  • rows[].verifiedboolean
    Has a domain or GitHub proof we verified
  • rows[].organicboolean
    Outside the wash registry and our confirmed clusters (our judgement)
  • rows[].servicestring | null
    The one model of a model-specific program; null for the provider's free models
  • rows[].paid_modelboolean
    The model is priced: the program is a free allowance, past it the buyer pays
  • rows[].models[]array
    Models the offer covers (one shared allowance)
  • rows[].program.per_addressinteger | null
    Free requests per buyer address per window without stake
  • rows[].program.per_ipinteger | null
    Free requests per IP per window without stake
  • rows[].program.window_msinteger
    The rolling window, ms (86,400,000 = a day)
  • rows[].program.stakers_onlyboolean
    Nothing free below the first tier
  • rows[].program.tiers[]array
  • rows[].program.tiers[].min_powerinteger
    Power the tier needs, ANTS-epochs
  • rows[].program.tiers[].multiplierinteger | null
    Times the base; null = unlimited
  • rows[].program.tiers[].requestsinteger | null
    Free requests per window at the tier, per stake; null = unlimited
  • rows[].program.tiers[].ants_one_yearinteger
    ANTS that reach it on a 52-week lock, rounded up
  • rows[].program.tiers[].ants_max_lockinteger
    ANTS that reach it with a max lock
  • rows[].youobject
    What the stake gets with this offer
  • rows[].you.tierinteger | null
    Tier the stake reaches (1-based); null = base or none
  • rows[].you.freeboolean
    Gets free requests (false: stakers only, below the first tier)
  • rows[].you.requestsinteger | null
    Free requests per window; null = unlimited
  • rows[].you.unlimitedboolean
    No cap
  • rows[].you.holds_weeksinteger | null
    Weeks the tier holds as the lock runs down; null with a max lock or no tier
  • rows[].you.requests_afterinteger | null
    Free requests per window once it drops; null = unlimited or no drop
  • rows[].you.next_tier_antsinteger | null
    ANTS to add, at the same lock, to reach the next tier; null at the top
  • rows[].you.next_tier_requestsinteger | null
    What the next tier gives; null = unlimited or none

Cache 45 s. Sources: program = announce, models = announce, name = announce, organic = antseedstats, you = antseedstats.

GET/staking/positionsEvery lANTS position

One row per staking position (the lANTS NFT) as the contract holds it: owner, pool, principal, stake, power this epoch and next, lock term, state. Filter by pool or state. Closed and withdrawn positions are included when asked for; the default is active plus pending.

curl "https://antseedstats.com/api/v1/staking/positions?state=open&limit=20"
/api/v1
Parameters
  • poolquery · int
    Only this pool (agent id)
  • statequery · open | active | pending | matured | closed | withdrawn | all · default open
    open: active plus pending (default); all: every position ever
  • ownerquery · address
    Only positions of this address
  • limitquery · int · default 100
    Rows to return (1 to 1000)
  • offsetquery · int · default 0
    Rows to skip
Response fields (24)
  • epochinteger
    Current epoch
  • totalinteger | null
    Positions matching; null when degraded
  • degradedboolean
    True when the market did not tie out
  • positions[]array
  • positions[].idinteger
    Position id (the lANTS token id)
  • positions[].ownerstring
    Lowercase 0x
  • positions[].owner_namestring | null
    Announced name when the owner is a seller (untrusted)
  • positions[].agent_idinteger
    Pool (agent id)
  • positions[].principal_antsnumber
    Put in
  • positions[].stake_antsnumber
    Counted as stake (principal plus restake bonus)
  • positions[].power_antsnumber
    Power this epoch
  • positions[].next_power_antsnumber
    Power next epoch
  • positions[].start_epochinteger
    First epoch the position counts
  • positions[].end_epochinteger
    Last epoch
  • positions[].term_epochsinteger
    Lock term as it applies in the epoch described
  • positions[].epochs_remaininginteger
    Left in the lock
  • positions[].max_lockedboolean
    At the maximum lock
  • positions[].originstring
    restaked, fresh or starter
  • positions[].statestring
    pending, active, matured, closed or withdrawn, relative to the current epoch
  • positions[].leaves_next_epochboolean
    Active now but not next epoch
  • positions[].pending_antsnumber
    Claimable now
  • positions[].has_nftboolean
    Still exists as the lANTS NFT
  • positions[].inherited_antsnumber
    Principal carried over from positions it replaced
  • positions[].replaced_next_epochboolean
    A child position carries it from the next epoch

Cache 5 to 8 min. Sources: positions = chain, owner_name = announce.

GET/rewards/currentThis epoch's budgets and top earners

The rewards page: the epoch's emission and bucket budgets as the gate reads them now, and the leading sellers and buyers by weighted usage points with the $ANTS each would receive if the epoch closed now (budget times share, capped per account). The projection is ours; the points and budgets are the contracts'. Everything settles at epoch close.

curl "https://antseedstats.com/api/v1/rewards/current?role=seller&limit=10"
/api/v1
Parameters
  • rolequery · seller | buyer | all · default all
    Which leaderboard to return
  • limitquery · int · default 10
    Rows per leaderboard
Response fields (41)
  • epoch.numberinteger
    Current epoch
  • epoch.starts_tsinteger
    UTC seconds
  • epoch.ends_tsinteger
    UTC seconds
  • epoch.seconds_leftinteger
    Until the next epoch
  • epoch.next_epoch_tsinteger
    UTC seconds
  • epoch.emission_antsnumber
    Emission for the epoch
  • epoch.modelstring
    legacy or recognized
  • degradedboolean
    True when a chain read failed; budgets, denominators, shares and projections are then null
  • budgetsobject | null
    Null when degraded
  • budgets.seller_usage_antsnumber
    Seller usage budget
  • budgets.buyer_usage_antsnumber
    Buyer usage budget
  • budgets.pools_antsnumber | null
    Staker pools budget
  • budgets.usage_input_microinteger
    usage input in micro-USDC (integer, 6 decimals)
  • budgets.usage_input_usdstring
    usage input in USDC, formatted with 2 decimals
  • budgets.usage_target_microinteger
    usage target in micro-USDC (integer, 6 decimals)
  • budgets.usage_target_usdstring
    usage target in USDC, formatted with 2 decimals
  • budgets.usage_pct_of_targetnumber
    %
  • budgets.stake_active_antsnumber
    Active stake
  • budgets.stake_target_antsnumber
    Target
  • budgets.stake_pct_of_targetnumber
    %
  • budgets.cap_bpsinteger
    Per-account cap, basis points
  • seller_weighted_pointsnumber | null
    Denominator for sellers; null when degraded
  • buyer_weighted_pointsnumber | null
    Denominator for buyers; null when degraded
  • sellers[]array
  • sellers[].accountstring
    Lowercase 0x
  • sellers[].namestring | null
    Announced name for sellers (untrusted)
  • sellers[].points_microinteger
    points in micro-USDC (integer, 6 decimals)
  • sellers[].points_usdstring
    points in USDC, formatted with 2 decimals
  • sellers[].weighted_pointsnumber
    Points times the pool weight, the contract's unit
  • sellers[].share_pctnumber | null
    Share of the side's weighted points, %; null when degraded
  • sellers[].cappedboolean | null
    Hit the per-account cap; null when degraded
  • sellers[].projected_antsnumber | null
    Budget times share, capped: our projection, settles at epoch close; null when the on-chain read is degraded
  • buyers[]array
  • buyers[].accountstring
    Lowercase 0x
  • buyers[].namestring | null
    Announced name for sellers (untrusted)
  • buyers[].points_microinteger
    points in micro-USDC (integer, 6 decimals)
  • buyers[].points_usdstring
    points in USDC, formatted with 2 decimals
  • buyers[].weighted_pointsnumber
    Points times the pool weight, the contract's unit
  • buyers[].share_pctnumber | null
    Share of the side's weighted points, %; null when degraded
  • buyers[].cappedboolean | null
    Hit the per-account cap; null when degraded
  • buyers[].projected_antsnumber | null
    Budget times share, capped: our projection, settles at epoch close; null when the on-chain read is degraded

Cache 45 s. Sources: epoch = chain, budgets = chain, points = chain, name = announce, projected_ants = antseedstats.

GET/rewards/poolsPools bucket, locked and staked totals

The rewards pools page: the per-seller pools cut tied out against the contract total (how many active sellers hold a pool, how many do not), the legacy locked rewards, the USDC seller stake and the claims to date. The seller counts are ours (indexed sellers asked of the contract); the totals are the contracts'.

curl "https://antseedstats.com/api/v1/rewards/pools"
/api/v1
Response fields (29)
  • epochinteger
    Epoch described
  • overviewobject
    Null when the gate did not read or the pools contract failed
  • overview.degradedboolean
    True when the numbers cannot be trusted
  • overview.reasonstring | null
    Why, when degraded
  • overview.known_sellersinteger
    Sellers we asked the pools contract about
  • overview.pools_with_stakeinteger
    Known sellers with active pool stake
  • overview.sum_pool_stake_antsnumber
    Sum of those positions
  • overview.unattributed_stake_antsnumber
    Contract total minus our sum
  • overview.active_sellersinteger
    Served inference this epoch
  • overview.active_with_poolinteger
    Of those, holding a pool
  • overview.active_no_pool_cleaninteger
    Of those, no pool and not a proven wash trader
  • overview.active_no_pool_washinteger
    No pool and in the registry
  • overview.clean_revenue_microinteger
    clean revenue in micro-USDC (integer, 6 decimals)
  • overview.clean_revenue_usdstring
    clean revenue in USDC, formatted with 2 decimals
  • overview.clean_settlesinteger
    Settlements of the clean no-pool group
  • overview.clean_revenue_share_pctnumber
    Their share of the epoch's non-wash volume, %
  • overview.no_pool_with_pointsinteger
    Sellers with points and no pool (should be 0)
  • locked.sellersinteger
    Sellers with locked legacy rewards
  • locked.total_antsnumber
    Locked, ANTS
  • locked.recordsinteger
    Lock records
  • stake.stakersinteger
    Sellers with USDC stake
  • stake.total_staked_microinteger
    total staked in micro-USDC (integer, 6 decimals)
  • stake.total_staked_usdstring
    total staked in USDC, formatted with 2 decimals
  • stake.total_slashed_microinteger
    total slashed in micro-USDC (integer, 6 decimals)
  • stake.total_slashed_usdstring
    total slashed in USDC, formatted with 2 decimals
  • claims.claimsinteger
    Claim events
  • claims.accountsinteger
    Distinct claimers
  • claims.total_antsnumber
    Claimed, ANTS
  • eligible_sellersinteger
    Sellers eligible for legacy claims

Cache 45 s. Sources: overview = chain, counts = antseedstats, locked = chain, stake = chain, claims = chain.

GET/referralsReferrers ranked: new buyers and ANTS

AIP-6 referrals. A referrer signs an invite, a new buyer carries it in the metadata it signs, and the binding is permanent. Each referred buyer's recognized usage (AIP-2 weighted points, wash already excluded on-chain) earns its referrer a share of the referrals bucket (2% of the epoch emission), split pro rata by points; during the first 12 epochs after binding the buyer earns the same amount as referee, so the referrer gets half of that buyer's share. ants_epoch is the open epoch's estimate and moves until it closes; ants_all_time sums every epoch. Points are the ledger's raw weighted points as strings (about 1e24 per USDC); use share, not points, to compare. live false means the contracts are not deployed yet.

curl "https://antseedstats.com/api/v1/referrals?limit=20"
/api/v1
Parameters
  • epochquery · int
    Epoch to rank by (default: the current one)
  • limitquery · int · default 100
    Rows to return (1 to 1000)
  • offsetquery · int · default 0
    Rows to skip
Response fields (18)
  • liveboolean
    The AIP-6 contracts are deployed and indexed
  • epochinteger
    Epoch ranked
  • buyers_referred_epochinteger
    Buyers bound to a referrer in this epoch
  • buyers_referred_all_timeinteger
    Buyers bound to a referrer, ever
  • active_referrersinteger
    Referrers that bound a buyer or earned points this epoch
  • pool_antsnumber | null
    The referrals bucket for the epoch, ANTS
  • totalinteger
    Referrers listed
  • referrers[]array
  • referrers[].rankinteger
    Position this epoch, 1-based
  • referrers[].addressstring
    Lowercase 0x
  • referrers[].namestring | null
    Announced seller name (untrusted)
  • referrers[].buyersinteger
    Buyers it brought, ever
  • referrers[].buyers_epochinteger
    Of those, bound this epoch
  • referrers[].pointsstring
    Ledger points this epoch (raw)
  • referrers[].sharenumber
    Share of this epoch's referral points, 0 to 1
  • referrers[].ants_epochnumber
    This epoch's ANTS, estimate while open
  • referrers[].ants_all_timenumber
    ANTS over every epoch
  • referrers[].claimed_antsnumber
    Claimed so far

Cache 45 s. Sources: referrers = chain, name = announce, ants_epoch = antseedstats, ants_all_time = antseedstats.

GET/referrals/{address}One wallet: rank, buyers, ANTS, invites

One wallet's referral standing. As referrer: rank this epoch, the buyers it brought (with the epoch each was bound and the last epoch that buyer earns its own referee share), ANTS this epoch and over all epochs, claimed, still claimable from closed epochs, this epoch's invite quota (read from the contract: 3 with any recognized activity last epoch, one more per 10 USDC, up to 20) and the operator rewards are paid to (null means none is set and claims revert until one is). As referee: who invited it and until which epoch it earns. by_epoch lists both roles. A wallet that never referred and was never referred answers with zeros.

curl "https://antseedstats.com/api/v1/referrals/0xa193c4005ede4705ea6b1582bfb45af1907e25ac"
/api/v1
Parameters
  • addresspath · address · required
    Wallet address (0x, 40 hex, any case)
Response fields (30)
  • liveboolean
    The AIP-6 contracts are deployed and indexed
  • addressstring
    Lowercase 0x
  • epochinteger
    Current epoch
  • as_referrer.rankinteger | null
    Position this epoch, 1-based
  • as_referrer.ants_epochnumber
    This epoch's ANTS, estimate
  • as_referrer.ants_all_timenumber
    ANTS over every epoch
  • as_referrer.claimed_antsnumber
    Claimed
  • as_referrer.claimable_antsnumber
    Closed epochs not claimed yet
  • as_referrer.invites_epochinteger | null
    Invite quota this epoch
  • as_referrer.invites_used_epochinteger
    Invites issued this epoch that bound a buyer
  • as_referrer.recipientstring | null
    Operator rewards are paid to; null when none is set
  • as_referrer.buyers[]array
  • as_referrer.buyers[].addressstring
    Lowercase 0x
  • as_referrer.buyers[].bound_epochinteger
    Epoch bound
  • as_referrer.buyers[].bonus_until_epochinteger
    Last epoch the buyer earns its referee share
  • as_referrer.buyers[].ants_all_timenumber
    ANTS this buyer earned its referrer
  • as_refereeobject
    Null when nobody referred this address
  • as_referee.referrerstring
    Lowercase 0x
  • as_referee.bound_epochinteger
    Epoch bound
  • as_referee.bonus_until_epochinteger
    Last epoch of the referee share
  • as_referee.ants_all_timenumber
    ANTS earned as referee
  • as_referee.claimed_antsnumber
    Claimed
  • as_referee.claimable_antsnumber
    Closed epochs not claimed yet
  • by_epoch[]array
  • by_epoch[].epochinteger
    Epoch
  • by_epoch[].rolestring
    referrer or referee
  • by_epoch[].pointsstring
    Ledger points (raw)
  • by_epoch[].total_pointsstring
    Epoch referral total (raw)
  • by_epoch[].antsnumber
    ANTS for the epoch
  • by_epoch[].claimedboolean
    Claimed

Cache 45 s. Sources: as_referrer = chain, as_referee = chain, invites_epoch = chain, ants = antseedstats.

GET/referrals/buildersBuilders: apps credited for usage

AIP-6 client attribution. The buyer's software puts an ERC-8004 agent id in the metadata it signs; that agent's owner earns a share of the builders bucket (4% of the epoch emission) pro rata to the recognized weighted points its buyers generate. Only usage naming a registered agent counts; attributed_share is how much of the epoch's usage did. The id is self-declared by the client, so a buyer can name its own agent: this is attribution, not proof of which product produced the usage.

curl "https://antseedstats.com/api/v1/referrals/builders"
/api/v1
Parameters
  • epochquery · int
    Epoch to rank by (default: the current one)
  • limitquery · int · default 100
    Rows to return (1 to 1000)
  • offsetquery · int · default 0
    Rows to skip
Response fields (15)
  • liveboolean
    The AIP-6 contracts are deployed and indexed
  • epochinteger
    Epoch ranked
  • pool_antsnumber | null
    The builders bucket for the epoch, ANTS
  • attributed_sharenumber | null
    Share of the epoch's recognized usage that named a registered app, 0 to 1
  • totalinteger
    Builders listed
  • builders[]array
  • builders[].rankinteger
    Position this epoch, 1-based
  • builders[].agent_idstring
    ERC-8004 agent id
  • builders[].namestring | null
    Announced name when the agent is also a seller (untrusted)
  • builders[].buyersinteger
    Distinct buyers credited, ever
  • builders[].pointsstring
    Ledger points this epoch (raw)
  • builders[].sharenumber
    Share of this epoch's attributed points, 0 to 1
  • builders[].ants_epochnumber
    This epoch's ANTS, estimate while open
  • builders[].ants_all_timenumber
    ANTS over every epoch
  • builders[].claimed_antsnumber
    Claimed so far

Cache 45 s. Sources: builders = chain, name = announce, ants_epoch = antseedstats, attributed_share = chain.

GET/diemDIEM pool: stake, yield, stakers

The DIEM provider-capacity pool: DIEM staked behind the Venice.ai proxy seller, the USDC it has paid stakers, the $ANTS allocated to them, the queued and unstaked amounts, and the pool's run-rate APY (our estimate from the last 7 days, or the last closed reward epoch). stakers lists the largest positions; series is DIEM staked over time. A figure whose on-chain read failed is null, never 0.

curl "https://antseedstats.com/api/v1/diem?limit=10"
/api/v1
Parameters
  • limitquery · int · default 50
    Stakers to return
Response fields (35)
  • proxy_sellerstring
    The DIEM staking contract, also the Venice.ai proxy seller
  • total_staked_diemnumber
    DIEM staked now
  • stakersinteger
    Stakers
  • usdc_distributed_microinteger
    usdc distributed in micro-USDC (integer, 6 decimals)
  • usdc_distributed_usdstring
    usdc distributed in USDC, formatted with 2 decimals
  • usdc_reserved_microinteger
    usdc reserved in micro-USDC (integer, 6 decimals)
  • usdc_reserved_usdstring
    usdc reserved in USDC, formatted with 2 decimals
  • ants_distributednumber | null
    $ANTS allocated to stakers (null when a reward-epoch read failed)
  • queued_unstake_diemnumber | null
    In open unstake batches (null when a batch read failed)
  • unstaked_diemnumber | null
    Flushed, not yet claimed (null when a batch read failed)
  • current_batchinteger
    Current unstake batch
  • reward_epochinteger
    Last finalized reward epoch
  • diem_price_usdnumber | null
    DIEM price used (null when the price chain is unavailable)
  • stake_value_usdnumber | null
    Stake value at that price
  • pool_yieldnumber | null
    Lifetime distributed over stake value
  • apy_pctnumber | null
    Annualised run-rate, %, our estimate
  • apy_basisstring | null
    rolling7d or epoch
  • rewards.usdc_microinteger
    usdc in micro-USDC (integer, 6 decimals)
  • rewards.usdc_usdstring
    usdc in USDC, formatted with 2 decimals
  • rewards.antsnumber
    $ANTS claimed
  • rewards.payoutsinteger
    Payout events
  • top_stakers[]array
  • top_stakers[].addressstring
    Lowercase 0x
  • top_stakers[].active_stake_diemnumber | null
    Staked (null when the read failed)
  • top_stakers[].queued_unstake_diemnumber | null
    Queued to unstake (null when the read failed)
  • top_stakers[].earned_usdc_microinteger | null
    earned usdc in micro-USDC (integer, 6 decimals)
  • top_stakers[].earned_usdc_usdstring | null
    earned usdc in USDC, formatted with 2 decimals
  • top_stakers[].usdc_claimed_microinteger
    usdc claimed in micro-USDC (integer, 6 decimals)
  • top_stakers[].usdc_claimed_usdstring
    usdc claimed in USDC, formatted with 2 decimals
  • top_stakers[].ants_claimednumber
    $ANTS claimed
  • top_stakers[].last_activity_tsinteger | null
    UTC seconds
  • series[]array
  • series[].tsinteger
    UTC seconds
  • series[].staked_diemnumber
    DIEM staked
  • series[].stakersinteger
    Stakers

Cache 45 s. Sources: economics = chain, apy = antseedstats, stakers = chain, series = chain.

GET/diem/proxyVenice.ai proxy: pool books and seller

The figures the /diem/proxy page shows, composed once per request from the same cached reads: the DIEM pool's accounting (unstake queue, reserved and claimed USDC, the pool's own reward epoch, the operator fee routed to buy and burn), its 50 largest stakers with every column, and the Venice.ai proxy seller's figures, daily and per-epoch volume and usage, its 100 most recent channels, 60 latest transactions and live offerings. A figure whose read failed is null, never 0: requests, tokens and offerings are null while the seller directory has no agent id for the proxy, and a day or epoch before one series' first record is null there. epochs are Antseed's 7-day epochs; pool.reward_epoch is the pool's own clock.

curl "https://antseedstats.com/api/v1/diem/proxy"
/api/v1
Response fields (81)
  • proxy_sellerstring
    The DIEM staking contract, also the Venice.ai proxy seller
  • agent_idstring | null
    The proxy's agent id (null while the seller directory has none)
  • pool.total_staked_diemnumber
    DIEM staked now
  • pool.stakersinteger
    Stakers
  • pool.queued_unstake_diemnumber | null
    In open unstake batches (null when a batch read failed)
  • pool.unstaked_diemnumber | null
    Flushed, not yet claimed (null when a batch read failed)
  • pool.current_batchinteger
    Current unstake batch
  • pool.oldest_unclaimed_batchinteger
    Oldest unclaimed unstake batch
  • pool.usdc_distributed_microinteger
    usdc distributed in micro-USDC (integer, 6 decimals)
  • pool.usdc_distributed_usdstring
    usdc distributed in USDC, formatted with 2 decimals
  • pool.usdc_reserved_microinteger
    usdc reserved in micro-USDC (integer, 6 decimals)
  • pool.usdc_reserved_usdstring
    usdc reserved in USDC, formatted with 2 decimals
  • pool.usdc_claimed_microinteger
    usdc claimed in micro-USDC (integer, 6 decimals)
  • pool.usdc_claimed_usdstring
    usdc claimed in USDC, formatted with 2 decimals
  • pool.usdc_payoutsinteger
    Payout events
  • pool.ants_distributednumber | null
    $ANTS allocated to stakers (null when a reward-epoch read failed)
  • pool.ants_claimednumber
    $ANTS claimed
  • pool.diem_price_usdnumber | null
    DIEM price used (null when the price chain is unavailable)
  • pool.stake_value_usdnumber | null
    Stake value at that price
  • pool.pool_yieldnumber | null
    Lifetime distributed over stake value
  • pool.apy_pctnumber | null
    Annualised run-rate, %, our estimate
  • pool.apy_basisstring | null
    rolling7d or epoch
  • pool.reward_epochinteger
    Last finalized reward epoch, the pool's own clock
  • pool.first_reward_epochinteger
    The pool's first reward epoch
  • pool.operator_fee_pctnumber
    Operator fee routed to buy and burn, %
  • pool.treasury_microinteger | null
    treasury in micro-USDC (integer, 6 decimals)
  • pool.treasury_usdstring | null
    treasury in USDC, formatted with 2 decimals
  • seller.earnings_microinteger | null
    earnings in micro-USDC (integer, 6 decimals)
  • seller.earnings_usdstring | null
    earnings in USDC, formatted with 2 decimals
  • seller.requestsinteger | null
    Requests served (null without an agent id)
  • seller.tokensnumber | null
    Input plus output tokens (null without an agent id)
  • seller.channelsinteger | null
    Channels settled at least once
  • seller.buyersinteger | null
    Distinct buyers
  • seller.offeringsinteger | null
    Live DHT listings (null without an agent id)
  • seller.stake_microinteger
    stake in micro-USDC (integer, 6 decimals)
  • seller.stake_usdstring
    stake in USDC, formatted with 2 decimals
  • seller.last_settle_tsinteger | null
    UTC seconds
  • top_stakers[]array
  • top_stakers[].addressstring
    Lowercase 0x
  • top_stakers[].active_stake_diemnumber | null
    Staked (null when the read failed)
  • top_stakers[].queued_unstake_diemnumber | null
    Queued to unstake (null when the read failed)
  • top_stakers[].earned_usdc_microinteger | null
    earned usdc in micro-USDC (integer, 6 decimals)
  • top_stakers[].earned_usdc_usdstring | null
    earned usdc in USDC, formatted with 2 decimals
  • top_stakers[].usdc_claimed_microinteger
    usdc claimed in micro-USDC (integer, 6 decimals)
  • top_stakers[].usdc_claimed_usdstring
    usdc claimed in USDC, formatted with 2 decimals
  • top_stakers[].ants_claimednumber
    $ANTS claimed
  • top_stakers[].last_activity_tsinteger | null
    UTC seconds
  • daily[]array
    One row per UTC day since the first settle or usage record
  • daily[].daystring
    UTC day, YYYY-MM-DD
  • daily[].settled_microinteger | null
    settled in micro-USDC (integer, 6 decimals)
  • daily[].settled_usdstring | null
    settled in USDC, formatted with 2 decimals
  • daily[].tokens_innumber | null
    Input tokens
  • daily[].tokens_outnumber | null
    Output tokens
  • epochs[]array
  • epochs[].epochinteger
    Antseed epoch
  • epochs[].settled_microinteger | null
    settled in micro-USDC (integer, 6 decimals)
  • epochs[].settled_usdstring | null
    settled in USDC, formatted with 2 decimals
  • epochs[].tokens_innumber | null
    Input tokens
  • epochs[].tokens_outnumber | null
    Output tokens
  • epochs[].requestsinteger | null
    Requests
  • epochs[].liveboolean
    The epoch still running
  • channels[]array
    The 100 most recently active
  • channels[].channel_idstring
    0x, 32 bytes
  • channels[].statusstring
    active, closed or unknown
  • channels[].buyerstring
    Lowercase 0x
  • channels[].settled_microinteger
    settled in micro-USDC (integer, 6 decimals)
  • channels[].settled_usdstring
    settled in USDC, formatted with 2 decimals
  • channels[].requestsinteger
    Requests
  • channels[].last_activity_tsinteger | null
    UTC seconds
  • transactions[]array
    The 60 latest; value is 0 for events that move no money
  • transactions[].tx_hashstring
    0x
  • transactions[].methodstring
    Decoded event
  • transactions[].blockinteger
    Block number
  • transactions[].value_microinteger
    value in micro-USDC (integer, 6 decimals)
  • transactions[].value_usdstring
    value in USDC, formatted with 2 decimals
  • transactions[].tsinteger | null
    UTC seconds
  • offerings[]array
  • offerings[].servicestring
    Model or service id
  • offerings[].input_usd_per_mnumber | null
    Input price per 1M tokens
  • offerings[].output_usd_per_mnumber | null
    Output price per 1M tokens
  • offerings[].tagsstring | null
    Comma-separated categories

Cache 45 s. Sources: pool = chain, apy = antseedstats, top_stakers = chain, seller = chain, daily = chain, epochs = chain, channels = chain, transactions = chain, offerings = announce.

Epochs and $ANTS
GET/epochsEvery epoch: emission, points, burn

Epochs are 7 days from genesis (2026-04-09 09:54:21 UTC). Points are the on-chain usage accruals in USDC; from epoch 22 they come from the usage accounting contract, which leaves out registry-proven wash traders, so totals drop at that boundary for the same real activity. burn_ants is what the gate burned at the close of that epoch.

curl "https://antseedstats.com/api/v1/epochs"
/api/v1
Response fields (13)
  • currentinteger
    Current epoch number
  • epochs[]array
  • epochs[].epochinteger
    Number
  • epochs[].starts_tsinteger
    UTC seconds
  • epochs[].ends_tsinteger
    UTC seconds
  • epochs[].statusstring
    complete, current or future
  • epochs[].emission_antsnumber
    Emission for the epoch
  • epochs[].modelstring
    legacy or recognized
  • epochs[].seller_points_microinteger | null
    seller points in micro-USDC (integer, 6 decimals)
  • epochs[].seller_points_usdstring | null
    seller points in USDC, formatted with 2 decimals
  • epochs[].buyer_points_microinteger | null
    buyer points in micro-USDC (integer, 6 decimals)
  • epochs[].buyer_points_usdstring | null
    buyer points in USDC, formatted with 2 decimals
  • epochs[].burn_antsnumber | null
    ANTS burned at close (null before the burn)

Cache 5 to 8 min. Sources: epochs = chain.

GET/epochs/{n}One epoch, live economics if current

Same row as /epochs. For the current epoch, economics adds the budget per bucket, the usage and stake inputs and the projected burn as read from the gate now. buyer_ants_per_usd_projected divides the buyer budget by buyer usage so far: our projection, it settles only at epoch close.

curl "https://antseedstats.com/api/v1/epochs/25"
/api/v1
Parameters
  • npath · int · required
    Epoch number
Response fields (41)
  • epochinteger
    Number
  • starts_tsinteger
    UTC seconds
  • ends_tsinteger
    UTC seconds
  • statusstring
    complete, current or future
  • emission_antsnumber
    Emission
  • modelstring
    legacy or recognized
  • seller_points_microinteger | null
    seller points in micro-USDC (integer, 6 decimals)
  • seller_points_usdstring | null
    seller points in USDC, formatted with 2 decimals
  • buyer_points_microinteger | null
    buyer points in micro-USDC (integer, 6 decimals)
  • buyer_points_usdstring | null
    buyer points in USDC, formatted with 2 decimals
  • burn_antsnumber | null
    Burned at close
  • economicsobject | null
    Current epoch only (null for past ones)
  • economics.minted_antsnumber | null
    Minted so far this epoch
  • economics.allocated_antsnumber | null
    Budgets granted
  • economics.unallocated_antsnumber | null
    Left over
  • economics.bucketsarray | null
  • economics.buckets[].keystring
    Bucket
  • economics.buckets[].cap_pctnumber
    Ceiling, % of emission
  • economics.buckets[].cap_antsnumber
    Ceiling
  • economics.buckets[].allocated_antsnumber
    Granted
  • economics.buckets[].dynamicboolean
    Scales with usage or stake
  • economics.burnobject | null
  • economics.burn.cap_antsnumber
    30% of emission
  • economics.burn.projected_antsnumber
    If someone settles
  • economics.burn.realized_antsnumber
    Burned so far
  • economics.usageobject | null
  • economics.usage.seller_microinteger
    seller in micro-USDC (integer, 6 decimals)
  • economics.usage.seller_usdstring
    seller in USDC, formatted with 2 decimals
  • economics.usage.buyer_microinteger
    buyer in micro-USDC (integer, 6 decimals)
  • economics.usage.buyer_usdstring
    buyer in USDC, formatted with 2 decimals
  • economics.usage.target_microinteger
    target in micro-USDC (integer, 6 decimals)
  • economics.usage.target_usdstring
    target in USDC, formatted with 2 decimals
  • economics.usage.pct_of_targetnumber
    Input over target, %
  • economics.usage.seller_budget_antsnumber
    Seller usage budget
  • economics.usage.buyer_budget_antsnumber
    Buyer usage budget
  • economics.stakeobject | null
  • economics.stake.active_antsnumber
    Active stake
  • economics.stake.target_antsnumber
    Target
  • economics.stake.pct_of_targetnumber
    %
  • economics.buyer_ants_per_usd_projectednumber | null
    Buyer budget over buyer usage so far
  • economics.degradedboolean
    True when a chain read failed: every figure in this block is then null

Cache 45 s. Sources: epoch = chain, economics = chain, buyer_ants_per_usd_projected = antseedstats.

GET/ants/supply$ANTS minted, emitted, burned, staked

The canonical token is 0xa87EE81b2C0Bc659307ca2D9ffdC38514DD85263 (transfers gated, no market price; an unrelated token with the same ticker is indexed elsewhere). emitted_ants counts completed epochs only. The partition is null when any identity it depends on failed to tie out.

curl "https://antseedstats.com/api/v1/ants/supply"
/api/v1
Response fields (19)
  • tokenstring
    Canonical $ANTS address
  • transfers_enabledboolean
    Token transfers open to everyone
  • total_supply_antsnumber
    totalSupply, includes burned-to-dead
  • max_supply_antsnumber | null
    Cap; null when unread
  • emitted_antsnumber
    Completed-epoch emission
  • burned_antsnumber | null
    Both burn sinks
  • burned_dead_antsnumber | null
    At 0x...dEaD; null when unread
  • partition.staked_antsnumber
    In pools
  • partition.stakable_antsnumber | null
    Staked plus earned not yet staked
  • partition.staked_pctnumber | null
    Staked over stakable
  • partition.minted_antsnumber
    Minted
  • partition.pre22_antsnumber
    Minted before epoch 22
  • partition.segments[]array
  • partition.segments[].keystring
    Segment
  • partition.segments[].antsnumber
    ANTS
  • partition.segments[].pctnumber
    % of total
  • burns[]array
  • burns[].epochinteger
    Epoch
  • burns[].antsnumber
    Burned at its close

Cache 5 to 8 min. Sources: supply = chain, partition = chain.

GET/ants$ANTS at a glance: supply, burn, holders

Everything the $ANTS page shows, composed once per request from the same cached reads. The canonical token is 0xa87EE81b2C0Bc659307ca2D9ffdC38514DD85263 (transfers gated, no market price). Net figures take the burn out: the epoch remainder is minted to the dead address, so it sits inside total_supply while held by nobody. A null is a read that failed, never a zero: a zero burn on a closed epoch is a zero read from the gate. where.slices are shares of the whole supply (minted plus rewards earned and not yet minted); the since-22 shares alone are on /ants/supply. split is the last closed epoch as the gate budgeted it, not what was paid. buy_burn.sent_to_zero_ants counts every transfer to the zero address, whoever sent it.

curl "https://antseedstats.com/api/v1/ants"
/api/v1
Response fields (60)
  • tokenstring
    Canonical $ANTS address
  • transfers_enabledboolean
    Token transfers open to everyone
  • epochinteger
    Live epoch
  • supply.total_supply_antsnumber
    totalSupply, the burn included
  • supply.max_supply_antsnumber | null
    MAX_SUPPLY
  • supply.emitted_antsnumber
    Completed-epoch emission
  • supply.burned_dead_antsnumber | null
    At 0x...dEaD: the epoch remainder burns
  • supply.minted_net_antsnumber | null
    totalSupply minus the burn
  • supply.minted_pct_of_maxnumber | null
    minted_net over max, %
  • supply.emitted_net_antsnumber | null
    Emission minus the burn
  • supply.minted_pct_of_emittednumber | null
    minted_net over emitted_net, %, capped at 100
  • holders.countinteger
    Addresses with a positive balance
  • holders.transfersinteger
    Transfer events indexed
  • holders.seriesarray | null
    Holders per day since the first transfer
  • holders.series[].daystring
    UTC day, YYYY-MM-DD
  • holders.series[].holdersinteger
    Positive balances at the end of that day
  • emission.per_epoch_antsnumber
    Scheduled emission of the live epoch
  • emission.next_halving.epochinteger
    First epoch of the next tier
  • emission.next_halving.starts_tsinteger
    UTC seconds
  • emission.next_halving.per_epoch_antsnumber
    Emission per epoch from then
  • emission.burned_epochsobject | null
  • emission.burned_epochs.burnedinteger
    Closed epochs that burned anything
  • emission.burned_epochs.closedinteger
    Closed epochs whose burn was read
  • emission.epochs[]array
    Epochs 22 to the live one
  • emission.epochs[].epochinteger
    Epoch
  • emission.epochs[].emission_antsnumber
    Scheduled emission
  • emission.epochs[].burned_antsnumber | null
    Burned at its close; 0 is a real zero; null unread or live
  • emission.epochs[].projected_burn_antsnumber | null
    Last closed epoch only, while its burn reads 0: what settling it would burn
  • emission.epochs[].liveboolean
    The live epoch: nothing minted yet
  • where.total_antsnumber | null
    Minted plus earned not yet minted
  • where.notestring | null
    Why the split is missing or coarser
  • where.slices[]array
    Empty when the split could not be shown
  • where.slices[].keystring
    pre22, burned, reserve, team, verification, wallets, residual, stakable, staked
  • where.slices[].labelstring
    Label
  • where.slices[].antsnumber
    ANTS
  • where.slices[].pctnumber
    % of total_ants
  • not_movable.legacy_escrow_antsnumber | null
    Still in the legacy-emissions escrow
  • not_movable.escrow_premint_antsnumber | null
    What funded the escrow
  • not_movable.locked_antsnumber | null
    Seller rewards locked in the pool
  • not_movable.locked_sellersinteger | null
    Sellers with a locked reward
  • not_movable.claims_openboolean | null
    A locked reward could be released now (policy installed and the pool allowed to move $ANTS)
  • splitobject | null
    Null when the gate could not be read
  • split.epochinteger
    Last closed epoch
  • split.emission_antsnumber
    Scheduled
  • split.allocated_antsnumber
    Sum of the budgets
  • split.remainder_antsnumber
    Emission no bucket was budgeted
  • split.minted_antsnumber
    Minted for it so far
  • split.burned_antsnumber
    Burned for it so far
  • split.buckets[]array
  • split.buckets[].keystring
    Bucket
  • split.buckets[].labelstring
    Label
  • split.buckets[].allocated_antsnumber
    Budgeted
  • split.buckets[].cap_antsnumber
    Ceiling
  • split.buckets[].dynamicboolean
    Scales with distance to target
  • buy_burnobject | null
    Null when the DIEM pool could not be read
  • buy_burn.treasurystring
    Operator fee recipient
  • buy_burn.fee_pctnumber
    Operator fee, %
  • buy_burn.treasury_microinteger | null
    treasury in micro-USDC (integer, 6 decimals)
  • buy_burn.treasury_usdstring | null
    treasury in USDC, formatted with 2 decimals
  • buy_burn.sent_to_zero_antsnumber
    $ANTS sent to the zero address, by anyone

Cache 45 s. Sources: supply = chain, emission = chain, where = chain, not_movable = chain, split = chain, holders = chain, buy_burn = chain.

GET/ants/holders$ANTS holders by balance

Balances rebuilt from every Transfer of the canonical token, protocol sinks (burn addresses) left out. The two largest holders are protocol contracts (the staking pools, the rewards contract); label names them. With transfers gated, a balance is rewards earned, not tokens bought.

curl "https://antseedstats.com/api/v1/ants/holders?limit=10"
/api/v1
Parameters
  • limitquery · int · default 30
    Holders to return
Response fields (8)
  • tokenstring
    Canonical $ANTS address
  • holdersinteger
    Addresses with a positive balance
  • transfersinteger
    Transfer events indexed
  • top[]array
  • top[].addressstring
    Lowercase 0x
  • top[].balance_antsnumber
    Balance
  • top[].labelstring | null
    Seller name or protocol role
  • top[].is_sellerboolean
    Has a seller page (not a protocol-only address)

Cache 5 to 8 min. Sources: holders = chain, label = announce.

Activity
GET/channelsPayment channels, most recent first

The channel lifecycle counts (opened, active, settled, closed) and the most recently active channels with what each has settled. A channel is one buyer paying one seller; it settles many times before it closes.

curl "https://antseedstats.com/api/v1/channels?limit=20"
/api/v1
Parameters
  • limitquery · int · default 100
    Channels to return
Response fields (14)
  • stats.openedinteger
    Channels ever opened
  • stats.activeinteger
    Open now
  • stats.settledinteger
    Settled at least once
  • stats.closedinteger
    Closed
  • channels[]array
  • channels[].channel_idstring
    0x, 64 hex
  • channels[].buyerstring
    Lowercase 0x
  • channels[].sellerstring
    Lowercase 0x
  • channels[].seller_namestring | null
    Announced name (untrusted); null when none
  • channels[].settled_microinteger
    settled in micro-USDC (integer, 6 decimals)
  • channels[].settled_usdstring
    settled in USDC, formatted with 2 decimals
  • channels[].settlesinteger
    Settlements
  • channels[].first_seen_tsinteger
    UTC seconds
  • channels[].last_activity_tsinteger
    UTC seconds

Cache 5 to 8 min. Sources: stats = chain, channels = chain.

GET/channels/{channelId}One channel: settlements and models

Everything the channel page shows: the pair, what it settled and when, the seller as the directory sees it, the per-model split when the buyer shares it, and the last settlements with their transactions.

curl "https://antseedstats.com/api/v1/channels/0x0000000000000000000000000000000000000000000000000000000000000001"
/api/v1
Parameters
  • channelIdpath · string · required
    Payment channel id (0x, 64 hex)
Response fields (32)
  • channel_idstring
    0x, 64 hex
  • buyerstring
    Lowercase 0x
  • sellerstring
    Lowercase 0x
  • settled_microinteger
    settled in micro-USDC (integer, 6 decimals)
  • settled_usdstring
    settled in USDC, formatted with 2 decimals
  • fees_microinteger
    fees in micro-USDC (integer, 6 decimals)
  • fees_usdstring
    fees in USDC, formatted with 2 decimals
  • settlesinteger
    Settlements
  • first_seen_tsinteger
    UTC seconds
  • last_activity_tsinteger
    UTC seconds
  • seller_infoobject
    Null when the seller is not in the directory
  • seller_info.namestring | null
    Announced name (untrusted)
  • seller_info.reputationnumber | null
    Trust score
  • seller_info.stake_usdcnumber
    USDC stake
  • seller_info.modelsinteger
    Models advertised
  • seller_info.regionstring | null
    Announced region
  • models[]array
  • models[].modelstring
    Canonical id
  • models[].model_displaystring
    Name
  • models[].databoolean
    A data service (a seller's own HTTP API billed per completed request), not a model: model is its announced service name
  • models[].settled_microinteger
    settled in micro-USDC (integer, 6 decimals)
  • models[].settled_usdstring
    settled in USDC, formatted with 2 decimals
  • models[].tokensinteger
    Tokens
  • models[].requestsinteger
    Requests
  • recent[]array
  • recent[].txstring
    Transaction hash
  • recent[].delta_microinteger
    delta in micro-USDC (integer, 6 decimals)
  • recent[].delta_usdstring
    delta in USDC, formatted with 2 decimals
  • recent[].total_settled_microinteger
    total settled in micro-USDC (integer, 6 decimals)
  • recent[].total_settled_usdstring
    total settled in USDC, formatted with 2 decimals
  • recent[].tsinteger
    UTC seconds
  • recent[].blockinteger
    Base block

Cache 45 s. Sources: channel = chain, seller = announce, models = chain.

GET/requestsLive requests: per-event usage reports

The usage reports decoded from settlements (MetadataRecorded) and free-tier reports, newest first: per event the model, tokens, requests, images and USDC, with the buyer, the seller, the channel and the transaction. window_24h is the rolling band the page opens with. Lower bounds: only buyers that share a per-model split appear.

curl "https://antseedstats.com/api/v1/requests?limit=20"
/api/v1
Parameters
  • limitquery · int · default 100
    Events to return
Response fields (46)
  • window_24h.sales_microinteger
    sales in micro-USDC (integer, 6 decimals)
  • window_24h.sales_usdstring
    sales in USDC, formatted with 2 decimals
  • window_24h.requestsinteger
    Paid requests
  • window_24h.tokensinteger
    Paid tokens
  • window_24h.imagesinteger
    Images
  • window_24h.video_secondsinteger
    Seconds of video
  • window_24h.videosinteger
    Videos delivered
  • window_24h.cached_sharenumber
    Cached over input, 0 to 1
  • window_24h.free_requestsinteger
    Free requests
  • window_24h.free_tokensinteger
    Free tokens
  • window_24h.buyersinteger
    Distinct buyers
  • window_24h.sellersinteger
    Distinct sellers
  • window_24h.modelsinteger
    Distinct models
  • window_24h.coveragenumber
    Settlements with a per-model split, 0 to 1
  • window_24h.top_by_tokens.modelstring
    Canonical id
  • window_24h.top_by_tokens.displaystring
    Name
  • window_24h.top_by_tokens.tokensinteger
    Tokens
  • window_24h.top_by_tokens.settled_microinteger
    settled in micro-USDC (integer, 6 decimals)
  • window_24h.top_by_tokens.settled_usdstring
    settled in USDC, formatted with 2 decimals
  • window_24h.top_by_tokens.requestsinteger
    Requests
  • window_24h.top_by_sales.modelstring
    Canonical id
  • window_24h.top_by_sales.displaystring
    Name
  • window_24h.top_by_sales.tokensinteger
    Tokens
  • window_24h.top_by_sales.settled_microinteger
    settled in micro-USDC (integer, 6 decimals)
  • window_24h.top_by_sales.settled_usdstring
    settled in USDC, formatted with 2 decimals
  • window_24h.top_by_sales.requestsinteger
    Requests
  • events[]array
  • events[].txstring
    Transaction hash
  • events[].blockinteger
    Base block
  • events[].tsinteger
    UTC seconds
  • events[].channel_idstring
    0x, 64 hex
  • events[].buyerstring
    Lowercase 0x
  • events[].sellerstring
    Lowercase 0x
  • events[].seller_namestring | null
    Announced name (untrusted); null when none
  • events[].modelstring
    Canonical id
  • events[].model_displaystring
    Name
  • events[].freeboolean
    Free-tier report, no payment
  • events[].requestsinteger
    Requests
  • events[].tokens_ininteger
    Input tokens
  • events[].tokens_cachedinteger
    Of which cached
  • events[].tokens_outinteger
    Output tokens
  • events[].imagesinteger
    Images
  • events[].video_secondsinteger
    Seconds of video
  • events[].settled_microinteger
    settled in micro-USDC (integer, 6 decimals)
  • events[].settled_usdstring
    settled in USDC, formatted with 2 decimals
  • events[].databoolean
    A data service (a seller's own HTTP API billed per completed request), not a model: model is its service name

Cache 45 s. Sources: window_24h = chain, events = chain.

GET/transactionsLatest on-chain activity, every kind

The most recent events across every Antseed contract we index: settlements, escrow deposits and withdrawals, USDC stakes, $ANTS claims, DIEM stakes and rewards, token transfers. One row per event with the account, the amount and its unit.

curl "https://antseedstats.com/api/v1/transactions?limit=20"
/api/v1
Parameters
  • kindquery · settle | deposit | withdraw | stake | unstake | claim | diem-stake | diem-unstake | diem-reward | transfer
    Only this kind of event
  • limitquery · int · default 100
    Events to return
Response fields (11)
  • events[]array
  • events[].txstring
    Transaction hash
  • events[].blockinteger
    Base block
  • events[].tsinteger
    UTC seconds
  • events[].kindstring
    settle, deposit, withdraw, stake, unstake, claim, diem-stake, diem-unstake, diem-reward or transfer
  • events[].accountstring
    Lowercase 0x
  • events[].account_namestring | null
    Announced name when the account is a seller (untrusted); null otherwise
  • events[].valuenumber
    Amount
  • events[].unitstring
    USDC, ANTS, DIEM or empty
  • events[].modelstring
    Model served (settle rows), empty otherwise
  • events[].model_idstring | null
    Canonical id when one model

Cache 45 s. Sources: events = chain.

Tools
GET/estimateCost of a workload on Antseed vs list

The calculator as an endpoint. Give a model and a monthly token mix; get the list price (OpenRouter, or Venice when OpenRouter has none or prices input at a sliver of output), the cheapest seller, the trusted pick (settled on-chain, trust score at or above 60, wash sellers never considered) and the saving. Costs include the 4% protocol fee; the list price does not. Image, video and audio generators are not priced (they bill per generation). Prices are advertised rates, not quotes.

curl "https://antseedstats.com/api/v1/estimate?model=kimi-k3&input_tokens=22000000&output_tokens=5500000"
/api/v1
Parameters
  • modelquery · string · required
    Canonical model id or any advertised spelling
  • input_tokensquery · int · default 22000000
    Input tokens per period (a coding agent: about 22 million a month)
  • output_tokensquery · int · default 5500000
    Output tokens per period
  • cached_tokensquery · int · default 0
    Cached input tokens per period (prompt prefixes already processed)
Response fields (39)
  • modelstring
    Canonical id
  • displaystring
    Name
  • tokens.inputinteger
    Input tokens
  • tokens.outputinteger
    Output tokens
  • tokens.cachedinteger
    Cached input tokens
  • fee_pctnumber
    Protocol fee applied to Antseed costs (4)
  • modestring
    full: list price and paid offers; market-only: no list price; free-only: every listing is free
  • listobject
    Null when the model has no published list price
  • list.input_usd_per_mnumber
    List USD per 1M input
  • list.output_usd_per_mnumber
    List USD per 1M output
  • list.cached_input_usd_per_mnumber | null
    List USD per 1M cached input
  • list.sourcestring
    Where the list price comes from
  • list.reference_idstring
    OpenRouter or Venice model id
  • list.cost_usdnumber
    Cost of the mix at list price, no fee
  • cheapestobject
    The cheapest paid seller for this mix; null when none
  • cheapest.sellerstring
    Lowercase 0x
  • cheapest.namestring
    Announced name (untrusted)
  • cheapest.input_usd_per_mnumber
    USD per 1M input
  • cheapest.output_usd_per_mnumber
    USD per 1M output
  • cheapest.cached_input_usd_per_mnumber | null
    USD per 1M cached input
  • cheapest.reputationnumber | null
    Trust score
  • cheapest.trustedboolean
    Passes the trusted rule
  • cheapest.verifiedboolean
    Domain or GitHub proof
  • cheapest.cost_usdnumber
    Cost of the mix, fee included
  • trusted_pickobject
    The cheapest trusted seller; null when none passes the rule
  • trusted_pick.sellerstring
    Lowercase 0x
  • trusted_pick.namestring
    Announced name (untrusted)
  • trusted_pick.input_usd_per_mnumber
    USD per 1M input
  • trusted_pick.output_usd_per_mnumber
    USD per 1M output
  • trusted_pick.cached_input_usd_per_mnumber | null
    USD per 1M cached input
  • trusted_pick.reputationnumber | null
    Trust score
  • trusted_pick.trustedboolean
    Always true
  • trusted_pick.verifiedboolean
    Domain or GitHub proof
  • trusted_pick.cost_usdnumber
    Cost of the mix, fee included
  • saving_usdnumber | null
    List cost minus the headline pick (trusted, else cheapest); null without a list price
  • saving_pctnumber | null
    The same as a percentage of list
  • free_sellersinteger
    Free listings of this model
  • free_trustedinteger
    Of those, passing the trusted rule
  • paid_sellersinteger
    Paid listings considered

Cache 90 s. Sources: list = antseedstats, prices = announce, trusted = antseedstats.

GET/glossaryEvery definition behind the site's tips

The text behind every info dot on the site as data: what each figure is and how it is computed, with the page it appears on. The same copy a visitor reads in the tooltip, so a figure and its definition can be cited together. Filter by page or search the text.

curl "https://antseedstats.com/api/v1/glossary?page=home"
/api/v1
Parameters
  • pagequery · string
    Only this page key (home, revenue, sellers, models, network, staking, rewards, ants, diem, buyers, calculator, status, developers, ...)
  • qquery · string
    Only definitions whose term or text contains this
Response fields (8)
  • totalinteger
    Definitions matching
  • terms[]array
  • terms[].idstring
    page.key, stable
  • terms[].termstring
    Human term
  • terms[].pagestring
    Page key
  • terms[].urlstring
    The page on antseedstats.com
  • terms[].definitionstring
    The definition, as shown in the tip
  • terms[].versionstring
    Definition version; bumps when the meaning changes

Cache never cached. Sources: definitions = antseedstats.

GET/changelogWhat changed on the site and the API

The site changelog as data, newest first: version, date, title and the items. An API change is always a changelog entry, so an agent can read what moved since it last looked.

curl "https://antseedstats.com/api/v1/changelog?limit=3"
/api/v1
Parameters
  • limitquery · int · default 10
    Entries to return
Response fields (7)
  • totalinteger
    Entries
  • entries[]array
  • entries[].versionstring
    Semver-like
  • entries[].datestring
    UTC day
  • entries[].titlestring
    Title
  • entries[].tagstring
    feature, fix or data
  • entries[].items[]array

Cache never cached. Sources: entries = antseedstats.

GET/usageAPI calls per day and client family

Calls to /api/v1 per UTC day, by client family (MCP, curl, Node, Python, browser, crawler, other), and the most called endpoints over the window. Counted per endpoint and family only: no IPs, no raw user agents. Today is the partial day so far. Includes our own checks and tests.

curl "https://antseedstats.com/api/v1/usage?days=7"
/api/v1
Parameters
  • daysquery · int · default 30
    UTC days to return, today included
Response fields (22)
  • days[]array
  • days[].daystring
    UTC day, YYYY-MM-DD
  • days[].callsinteger
    Calls that day
  • days[].mcpinteger
    Calls from our MCP server
  • days[].curlinteger
    curl, wget, httpie
  • days[].nodeinteger
    Node, Bun, Deno
  • days[].pythoninteger
    Python clients
  • days[].browserinteger
    Browsers
  • days[].botinteger
    Crawlers
  • days[].otherinteger
    Unrecognised user agents
  • totalsobject
    The whole window
  • totals.callsinteger
    Calls in the window
  • totals.mcpinteger
    Calls from our MCP server
  • totals.curlinteger
    curl, wget, httpie
  • totals.nodeinteger
    Node, Bun, Deno
  • totals.pythoninteger
    Python clients
  • totals.browserinteger
    Browsers
  • totals.botinteger
    Crawlers
  • totals.otherinteger
    Unrecognised user agents
  • endpoints[]array
  • endpoints[].idstring
    Endpoint id
  • endpoints[].callsinteger
    Calls in the window

Cache 45 s. Sources: days = antseedstats, totals = antseedstats, endpoints = antseedstats.

Health
GET/statusIndexer head, lag, subsystems

Never cached. last_indexed_block is the oldest watermark across the ingestion streams, which is also the as_of_block every other endpoint reports.

curl "https://antseedstats.com/api/v1/status"
/api/v1
Response fields (13)
  • overallstring
    ok, warn or down
  • head_blockinteger | null
    Base head, read live
  • last_indexed_blockinteger | null
    Oldest stream watermark
  • lag_blocksinteger | null
    head minus last indexed
  • streams[]array
  • streams[].keystring
    Stream
  • streams[].last_blockinteger
    Watermark
  • streams[].lag_blocksinteger | null
    Lag
  • streams[].statusstring
    ok, warn or down
  • subsystems[]array
  • subsystems[].keystring
    Subsystem
  • subsystems[].statusstring
    ok, warn or down
  • subsystems[].detailstring
    Human summary

Cache never cached. Sources: status = antseedstats.

Built for agents too: llms.txt is the front door, llms-full.txt the whole backstage in one file, and For agents explains the pattern. Data under CC BY 4.0, attribute to AntSeedStats (antseedstats.com).

Built by @gekko_eth. Missing an endpoint or a tool? Ask there.