KertivoWhat does a Kertivo Index grade mean (A / B / F / went-dark)?
A Kertivo Index grade is the result of one dated paid test: we send real Base USDC to the endpoint and score whether settlement and delivery matched what the seller advertises. A means settled + HTTP 200 + schema match within the declared timeout; B means settled + 200 with a partial schema or slow response; F means money moved without a matching 200 (no 200 at all, or a 200 with none of the advertised fields); C-went-dark / C-never-live mean nothing delivered and nothing moved. Grades are earned, never bought.
Who this is for
- Agents reading an A/B/F on /trust-index or a
/trust/lookupresponse and needing the letter definitions. - Sellers who want to know what an Index row actually measured before they cite it.
- Auditors comparing marketing claims to an independent USDC payment receipt.
How a grade is earned (methodology)
Each weekly Index run pays listed Base USDC x402 endpoints from a capped test wallet (Kertivo Index #4: global cap $0.50, per-call cap $0.01, payer 0x9D42EbA8868CFd38A50Ba615ee0cf903F1BBeCE9). For every row we record the unpaid 402 (price, payTo, network), the paid settlement, the HTTP status after pay, latency vs the seller's declared timeout, and whether the body matches the seller's declared output example, path by path up to three levels deep: at least 80% of the example's fields present is a match, some but fewer is partial, none is a mismatch. If the seller declared no example, any non-empty JSON body counts as a match.
Scorecard for Kertivo Index #4 (generated Oct 25, 2026, 7:37 PM PDT / 2026-10-26T02:37:43.703Z), 116 endpoints: A = 95, B = 7, F = 3, C-went-dark = 4, C-never-live = 6, ungraded = 1, total spent 0.418801 USDC, settled payments 105, rubric v3. Machine-readable: latest.json, raw.jsonl. Frozen past runs stay at /trust-index/runs.
Each weekly run publishes new dated counts, and some letters change; this guide's numbers are from Kertivo Index #4, so check the live scorecard for the current run. Letter meanings are the same in every run (A / B / F / C-went-dark / C-never-live).
Grade letter table
| Grade | Meaning |
|---|---|
| A | Settled on-chain + HTTP 200 + schema match + within declared timeout |
| B | Settled + 200, but partial schema or slow |
| F | Money moved without a matching 200: no 200 at all, or a 200 with none of the advertised fields (however fast) |
| C-went-dark | Nothing delivered, nothing moved — and this endpoint did deliver in an earlier run |
| C-never-live | Nothing delivered, nothing moved — and we have no earlier successful delivery on record |
| ungraded | In this run's list but not paid (over the per-call price cap or skipped by policy), so no letter |
What a grade proves (on the run date shown): we paid the listed price in real Base USDC, the payment settled, and — for A/B — the endpoint returned a valid response in the shape it advertises.
What it does not prove: that the seller is honest, that the data is accurate, or that the endpoint will keep working after the test. A grade is one dated paid test, not an endorsement.
Grades are earned, never bought. Sellers cannot pay Kertivo for an A. Badges (GET /badge/<url-encoded-endpoint>) show the latest Index letter (or "not graded"); they are earned, never sold.
Claims vs grades
Buyer-filed claims sit next to a grade and never rewrite it. A claim proves an on-chain payment; "disputed" vs "delivered" is the buyer's report. Sellers can answer with a signed reply. Claims on unindexed URLs still exist; they never become a grade.
Our own Kertivo routes, when included in a run, are graded by the same rubric. Self-tests are disclosed; we do not hand-adjust our rows. In Index #4 that gave /check, /trust/lookup, /address/check and /address/check/full an A (graded on our workers.dev address, the same service with the same price and payTo).
Try it
- Read the live scorecard: https://kertivo.com/trust-index.
- Free one-URL grade:
curl -s 'https://kertivo.com/trust/lookup?url=https://api.onesource.io/api/chain/erc20-balance'→ expectstatus: indexed, a letter, andrunDate. - Unknown URL: the same GET with a non-indexed URL returns
status: not_indexed,grade: null— no invented score. - Paid evidence ($0.002):
POST /trust/lookupwith{url}after the unpaid 402. Shapes on /examples.
FAQ
Can a seller buy an A?
No. The scorecard states grades are earned by settling a real Base USDC payment and delivering in the paid test.
Why is something B instead of A?
It settled and returned 200, but either only some of the advertised fields were in the body, or the response came back slower than the seller's declared timeout (maxTimeoutSeconds from its 402; 5 seconds if it declares none). A 200 with none of the advertised fields is F, not B, however fast or slow it was.
What is the difference between C-went-dark and C-never-live?
Both mean this run moved no money and got no delivery. Went-dark means an earlier Index run did see delivery; never-live means we have no such earlier success on record.
Is lookup a live re-pay of the seller?
No. Free GET and paid POST /trust/lookup read the latest published Index (plus claims). They do not re-pay the seller on every request.
What does "ungraded" mean?
The endpoint was in this run's list, but we did not pay it, usually because its price is above the run's per-call cap. It says nothing good or bad about the seller.
Will my grade change next week?
It can. Each run is dated. Older frozen runs remain at /trust-index/runs. Badges follow the latest published letter.
You graded yourselves. Is that fair?
When our routes are in the selection, payment goes from the test wallet to our business wallet and is labeled as a self-test. Same rubric, no special overrides for our rows.
Limits
- A grade covers one paid test on one run date — not continuous monitoring.
- We only grade endpoints that appear in a published Index selection, not every URL on the web.
- Schema checks follow the seller's advertised example / paths; they do not prove semantic correctness of the data.
- Deep-priced routes above the per-call Index cap (for example
/check/deepat $0.05) are out of a capped weekly run. - Claims never rewrite grades; grades never imply arbitration or custody.
- Kertivo holds no funds and does not arbitrate disputes. It reports evidence.