Reputation

Buyers score sellers locally from data they can check themselves: settled service history, the seller pool's share of recognized usage and staking power, and wash-trading verdicts, all read from the chain, plus public history of identities the seller has proven it owns. There is no central reputation authority and no seller allowlist.

Trust score​

Every buyer computes one number per seller, 0-100, as a weighted sum of independent parts:

trust = washFlagged ? 0 : history + usage + power + identity

Each part is a 0-1 value times its weight. History averages log10(1 + channelCount) / log10(101) and log10(1 + settledVolumeUsdc) / log10(101), with each term capped at 1. Both network shares are unitless and go through the same log curve, log10(1 + 999 · share) / 3: a 100% share maps to 1, 10% to 0.67, and 1% to 0.35, so shares self-normalize as the network grows or as more ANTS is staked. Identity maps its points (GitHub up to 70, domain up to 12, below) onto 0-1 over 70. The weights sum to 100 and live in one table (TRUST_WEIGHTS); a future model-verification part will take its weight from there.

The history part only needs the channels and seller-registry contracts already required for paid routing. The other on-chain parts need the recognized-usage stack: sellerPoolsAddress, usageAccountingAddress, and washTradingRegistryAddress in the chain config, filled automatically for base-mainnet. On chains without the recognized-usage stack, service history and identity can still score a peer; in the first epoch after activation there is no previous-epoch usage share yet. Buyers read every on-chain input for a discovery pass in two Multicall3 round trips (chunked at 80 calls each) and refresh a seller at most every 120 seconds.

Identity​

The identity part is the strongest single verified identity; several accounts or domains owned by one operator never add up.

• GitHub portfolio (max 70). Original, non-fork repositories at least three months old with at least five stars, excluding the ownership-proof repository: 40 · min(1, Σ log2(1 + min(stars, 500)) / 40) for stars, + 20 · min(1, projects / 8) for breadth, + 10 · min(1, oldest project years / 3) for age. Archived repositories count at 20%. An empty or zero-star account earns 0.
• Domain registration age (max 12). 12 · min(1, years / 5), read from the authoritative RDAP registry. The registration must match the exact domain, so app.example.com cannot inherit the age of example.com.
• Collection. Only identities whose ownership proof verified are looked up. GitHub: one account lookup plus up to four pages of 100 repositories, keyed by numeric account id so a reassigned username cannot inherit the previous owner's history. Domains: the cached IANA RDAP bootstrap plus one registry query. Requests go only to public HTTPS hosts, do not follow redirects, time out after 8 seconds, and are capped at 2 MB. Evidence is buyer-local, usable for seven days, and never part of signed seller metadata.
• Public history is a heuristic, not proof of service quality: stars can be bought, and accounts and domains change hands.

Not in the score​

These stay separate router rules and never change the number: the local sybil heuristic (a display-only warning in the CLI and desktop), ghost-channel count, settlement recency, failure streaks and cooldowns, price limits, and allow/block lists. A seller-reported reputationScore is used only when the buyer has not scored the peer.

Routing​

The buyer proxy and desktop share the route ranking exported by @antseed/node/model-routing, and /v1/models/:id returns peer offers in that same order. The default model routing preferences are:

{
  preferFreePeers: false,
  maxInputUsdPerMillion: 25,
  minTrustScore: 60,
  allowedPeerIds: [],
  blockedPeerIds: [],
}

minTrustScore and the allow/block lists are hard eligibility rules. At the default 60, unscored sellers are excluded. An established seller that has reached both history targets scores 50 before usage, power, or identity; a 10% power share adds about 6.7 points and a 1% share of last epoch's recognized usage about 7, so demonstrated service alone does not pass the default gate: it also takes recognized usage, pool participation, or a verified identity. Identity alone reaches at most 20. Buyers can lower buyer.routingPreferences.minTrustScore, or set it to 0 to consider unscored peers. buyer.minPeerReputation and hierarchical maxPricing remain separate hard policy checks applied before the ranking.

Eligible offers are ranked by trust, token or image price, cached-input pricing coverage, free-peer preference, recent failures, and cooldown state. If at least one seller for a model advertises cached-input pricing, offers that omit it receive a model-specific reputation reduction; if none advertise it, no seller is penalized. A recognized conversation softly prefers its previous successful seller while that offer remains healthy and eligible. Latency is tracked as an exponential moving average (alpha: 0.3), and peers with consecutive failures enter exponential backoff cooldown.

The lower-level @antseed/router-core package also exposes generic router weights for plugin authors; its reputation factor is the trust score when available:

On-Chain Stats​

AntSeed derives core on-chain seller stats directly from AntseedChannels. Completed channels, ghost channels, and settled volume live in the Channels contract itself. An optional AntseedStats contract can additionally ingest buyer-signed metadata during settlement to aggregate token and request counters. Buyers still read these counters for display and for the local sybil warning; they are not part of the trust score.

Each seller's ERC-8004 agentId maintains the following core counters in AntseedChannels:

If the optional AntseedStats contract is configured, it can also track:

No counter can be incremented without a corresponding on-chain state transition and buyer-signed metadata hash.

Staking​

From September 10, 2026 at 09:54:21 UTC (epoch 22), seller eligibility is resolved through AntseedSellerRegistry and ANTS pool positions contribute epoch power. Legacy USDC stake can remain an eligibility fallback while enabled; recognized-usage rewards require pool power, and the same pool power share is the power part of the trust score. See Recognized Usage and legacy USDC staking.

ERC-8004 Feedback​

Buyers submit structured feedback via the deployed ERC-8004 ReputationRegistry (Base: 0x8004BAa1...). Feedback signals:

Feedback and the trust score do not automatically grant ANTS or apply a fixed on-chain reward multiplier. Recognized-usage accounting applies the configured reward policies; the registered historical wash-trading policy can zero future rewards without changing settlement stats.

ANTS Rewards​

Protocol start: September 10, 2026 at 09:54:21 UTC (epoch 22).

The trust score helps a buyer choose a seller. ANTS reward accounting is separate: it uses recognized usage, seller-pool power, and the configured policies. A high trust score does not override a reward exclusion.

See Recognized Usage and ANTS Rewards for the standard reward model. Looking for pre-migration emissions? The legacy guide covers the 65% seller / 5% buyer split and historical claims.