1. 1

   Discovery

   New endpoints arrive continuously from on-chain settlement signals across every supported chain, from facilitator catalogs, and from merchant claims.

   SolanaBasePolygonArbitrumOptimismAvalancheBNBSKALE
2. 2

   Classification

   Each endpoint gets probed before any real call goes out. Endpoints that return a valid 402 with payment requirements move forward. Endpoints that 5xx, time out, or return 200 without asking for payment get skipped.

3. 3

   Verification

   Eligible endpoints get an input that fits the endpoint, the call gets made, and the full response is captured along with the settlement record.

4. 4

   Grading

   The response is evaluated against what the merchant claims the endpoint does, including the URL, description, price, and the input we sent. The output is a 0 to 100 score with a status, brief notes, and concrete fix instructions if something is broken.

5. 5

   Publish or suspend

   High-scoring endpoints surface in search, ranked by a blend of score, settlement volume, and recency. Endpoints that fail repeatedly are auto-suspended and stop appearing in default results.

6. 6

   Backoff retry

   Suspended endpoints re-enter the queue on a widening backoff schedule. A single passing run resets the counter and restores the listing. Endpoints that stay broken long enough are eventually retired, and an admin can revive them later.

7. 7

   Merchant overrides

   Merchants can run a test on their own endpoint at any time from the resource editor. A passing run resets the failure counter and lifts suspension immediately.

Every test produces a 0 to 100 score plus a status of pass, fail, or inconclusive. The AI judges whether the response actually delivers on the merchant's claimed purpose, given the price paid.

90–100

Excellent

The response does what the endpoint says it does, and the structure of what comes back is easy to consume. Surfaces high in search.

75–89

Good

The response works. Documentation might be sparse or the shape inconsistent, but the listing is reliable enough to keep.

50–74

Mediocre

The response works, but it does not quite match the description. Listed, but ranked lower until the merchant tightens it up.

0–49

Failing

Something is wrong. The response is broken, the price is wrong for what comes back, or the description is misleading. Hidden from default search.

Badges high-scoring listings can earn

90+ Club

Top resource

Top server

Bulletproof

Across the stack

Wash-free

Every row in the verification history is tagged so merchants and admins can audit who paid for what.