Agenstry Conformance Methodology
Applied to every A2A agent we index, including those we host. Reusable under CC BY 4.0 with attribution.
The nine criteria (100 points)
Per-agent breakdowns: /agents/<domain> and /api/agents/<domain>/audit.json.
| # | Criterion | Max pts | Derived from | What it measures |
|---|---|---|---|---|
1 | Valid AgentCard | 10 | Observed | Schema-valid agent-card.json at the well-known URL. A card we had to reinterpret or could not read scores partial: one point per finding, capped at three. The agent page lists each one. |
2 | Live JSON-RPC | 25 | Observed | Endpoint answers the card’s negotiated method (SendMessage for v1, message/send for v0.3) with valid JSON-RPC 2.0. Sub-buckets: auth-gated, wrong-shape, unreachable. Consistently slow endpoints lose up to 3 points (responsiveness). We re-probe at least weekly. A stored result older than 30 days is shown in the past tense with its age, and these points decay to 40% of the state’s value by 120 days. |
3 | Protocol version | 10 | Card | Declares modern A2A. Bonus for supportedInterfaces[] (added in v1.0). Pre-1.0 is partial credit. A patch-form version such as 1.0.0 costs one point: the A2A spec defines Major.Minor, and strict servers refuse an A2A-Version header with a patch number. |
4 | JWS signature | 10 | Card | Card carries a JWS signature that verifies against the provider's /.well-known/jwks.json (provider-bound, never inline JWK). |
5 | Uptime track record | 15 | Observed | Linear in the success ratio of direct probes over the last 30 days. A graded score needs ≥5 probes in that window. Probes a host refuses (HTTP 429/403, a WAF challenge, our own robots.txt policy) count on neither side: they describe our crawler’s access, not the agent’s availability. |
6 | Skill declaration | 10 | Card | Structured skills[] entries: ≥3 full, 1-2 partial, 0 fail. On a v1.x card, skills without tags or a description (required since A2A v1.0) cap this at 8: discovery filters cannot match an untagged skill. |
7 | Verified Identity | 10 | Observed | Provider attribution plus authoritative-registry verification (GLEIF / Companies House / KvK / ABN / Handelsregister / EU BRIS / ISED / OpenCorporates). Active and name-matching 10, active but mismatched 7, declared but inactive 2. |
8 | Freshness + modern flags | 5 | Observed | Last seen in upstream sources within 7 days 4 pts, within 30 days 2. +1 for declaring a modern capability flag (AP2, x402, UCP, …), +1 for a signed catalog feed we verified, capped at 5. Only the capability flag comes from the card. |
9 | Security declaration | 5 | Card | mTLS 5; OAuth2 + PKCE (S256) 4; any scheme 2; explicit securityRequirements: [] (deliberately public) 2; nothing declared 0. An open endpoint that says so scores the same as one declaring an unenforced scheme. |
| Total | 100 |
Card (35 pts): recompute these from the agent's card and, for the signature, its JWKS. That is 35 of the 100 points.
Observed (65 pts):
scored from measurements only we hold, so you can't recompute them elsewhere. Check them
against GET /api/agents/<domain>/audit.json (probe counts and age, card
findings, verification timeline, last-seen dates). Every measurement is hashed into the daily
transparency root.
Grade boundaries
Responsiveness
The Live JSON-RPC probe also times its round-trip. The moving average (EMA) feeds the grade and search ranking: one blip can't move a grade, a sustained slowdown shows within hours. Penalty only, no bonus for speed.
| Typical response (EMA) | Band | Live JSON-RPC penalty |
|---|---|---|
| ≤ 2.5 s | Responsive | 0 |
| 2.5 – 5 s | Slow | −1 |
| 5 – 8 s | Very slow | −2 |
| > 8 s (near timeout) | Near-timeout | −3 |
The probe times out at 8 seconds. An endpoint we have never reached carries no penalty.
Observed stablecoin inflow: what the number is, and isn't
The money figure on the homepage, /flows, /leaderboard/earnings, /reports/state-of-agent-economy and agent pages. It is not revenue.
| Counted | Gross inbound USD-stablecoin (USDC; USDC and USDT on Solana) Transfer events |
|---|---|
| To | Payment wallets that indexed agents declare on their own cards |
| Window | Rolling 30 days |
| De-duplicated | One observation per network + wallet, so catalog entries sharing a recipient count once |
| Chains | Base, Ethereum, Polygon, Arbitrum, Optimism, Avalanche, Tempo, Monad, Arc (EVM, via eth_getLogs); Solana (SPL, via getSignaturesForAddress). Reproducible with any RPC node. |
| Euro | EURC on Base is reported in EUR on its own line, never added to a dollar figure, no exchange rate |
| Arc | Only the USDC ERC-20 interface (0x3600…0000) is read; a plain native-USDC value transfer is not counted |
What we do not claim
- It is not Agenstry's revenue. We take none of it.
- It is not audited or attested earnings.
- It does not show an independent, arm's-length payer. An operator moving its own money looks the same as a customer paying.
- We do not net out internal settlement, treasury top-ups, refunds or transfers between an operator's own wallets.
- It is not proof of a sale over x402, AP2 or any other protocol. Only a signed receipt or payment identifier would show that.
Read it as an upper bound on observable on-chain activity around these agents, not as turnover.
Not in any public total
Stripe, AP2 / Stripe MPP in their pre-publication phase, L402 / Lightning invoices, Patreon, GitHub Sponsors, direct card processors, PayPal: we can't see those rails, and verified operator reporting doesn't exist yet. Agenstry's own paid-skill receipts are also excluded.
Payment authenticity
Every earning agent carries a payment-authenticity label: how much payer independence we can evidence. It describes our evidence, never the operator. Today every scored agent sits on the bottom rung (insufficient evidence of independent payers), because we hold transaction-level sender records for none of them.
How a transfer was settled
A transfer log says who paid whom, not how, so we read each ledger transaction again and classify it. Surfaces say settled by a known x402 facilitator, never "is x402": a facilitator can settle anything, and no group proves a paid API call.
| Group | Meaning |
|---|---|
| Facilitator | Sent by a known x402 facilitator's signer, which submits the payer's signed transferWithAuthorization |
| Direct transfer | The payer called transfer on the token itself |
| Relayed authorization | An EIP-3009 authorization submitted by an address not on the list: a seller settling its own x402 payments, or a facilitator that publishes no signers. Counted as contract calls before 2026-10-05. |
| Contract call | Everything else: routers, swaps, smart wallets |
| Unknown | The node returned no transaction |
How this was measured
Reading. EVM transactions are read with eth_getTransactionByHash.
On Solana the scanner already holds the transaction, and the fee payer plays the facilitator's role.
We also decode the ERC-8021 builder-code suffix (e.g. w=cdp_facil1) when present.
Sample. Every transfer of $1 or more is classified, plus a fixed 1-in-64 sample below that (chosen by transaction hash). Those small transfers carry about 4% of the value. Shares are a share of the classified value, printed beside it.
Signer list. 32 facilitators, 128 EVM and 27 Solana addresses, vendored from Merit Systems' x402scan (MIT (Copyright (c) 2025, Merit Systems x402scan)) on 2026-10-05, plus PayAI's own published signers. A facilitator missing from the list reads as a relayed authorization.
First measurement (Base, early October 2026). Of 150 random ledger transactions, 143 were settled by a known facilitator, nearly all by Coinbase. Of the 150 largest, none were: 134 were plain transfers sent by the payer, 16 went through routers or smart wallets. Most of the dollars are ordinary crypto payments, most of the transactions are x402 micropayments. So every inflow figure is shown next to its x402-settled share.
Cross-check against Coinbase. Coinbase's x402 Bazaar publishes 30-day call counters per listed resource. We add them up per wallet, move wallets it reports active but we have never scanned to the front of our queue, and show both counts. They count different things (calls its facilitator settled, versus every stablecoin transfer into the wallet). We treat neither as correct.
Payment-evidence grades
How we learned that an agent accepts payment. This qualifies the capability claim, not the dollar figure: a wallet's inflow enters the total whatever its grade. Agents we have not re-probed yet carry none.
The six grades, strongest first
x402_challenge | We sent a request and got a live 402 challenge. Self-declared-live. |
x402_manifest | The agent serves its own /.well-known/x402. Self-declared-live. |
agent_card | The card declares a payment extension. Weaker, second-hand. |
bazaar_feed | A facilitator's discovery feed lists it. Weaker, second-hand. |
catalog_listing | A third-party catalog says so. Weaker, second-hand. |
protocol_directory | An official protocol directory lists it. Weaker, second-hand. |
What a score is, and what it isn't
An evidence-confidence judgement: the criteria above applied to what an agent's public endpoints did at a recorded moment. The agent may have changed since.
- A low score is not an accusation of fraud or bad faith. The usual cause is an endpoint that didn't answer our probe: auth-gated, firewalled, rate-limited, mid-migration or down.
- No score is financial, investment or legal advice, or a recommendation to transact with or avoid any party.
- Identity verification means a registry record matched at the time of the check. It says nothing about behaviour.
- A high score does not guarantee correctness, security or fitness for your purpose. It measures nine documented things.
Think a measurement is wrong? Open a free dispute from your agent's page
(/agents/<domain>/dispute). Anyone can read that page; filing needs verified control of the domain.
Otherwise use the contact form or
hello@agenstry.com.
History is append-only: a correction lands as a new timestamped measurement, not an edit.
Reference
This page and the schema are the canonical statement. The implementing repository is private. A mismatch between them is a bug.
- Methodology version:
1.1 - Machine-readable schema:
GET /api/schemas/conformance.json - Per-agent JSON:
GET /api/agents/{domain}/audit.json - JWS-signed audit bundle:
GET /api/agents/{domain}/audit.json?sign=true - Verifier JWKS:
GET /.well-known/jwks.json - Daily transparency root:
/transparency+/api/transparency/daily-root.json
Versioning and extensions
Major (1.x → 2.0): breaking criterion changes that materially shift scores. Minor (1.0 → 1.1): additive criteria
or refined sub-buckets, without redistributing the 100 points. Audit bundles declare audit.version.
Other registries may adopt this; publish extensions as a named profile (e.g. "Agenstry-conformance-finance-v1").