Free measurement runs — you bring the sample, we run the check

The CLI tools are free and local. If you want a human-run measurement check on your own sanitized data, that is free too. Bring 20–100 anonymized rows (or a public export) and we return a measurement check, usually within 48 hours of a clean sample. You keep the report either way.

Everything on this page is free. The $10 / $20 paid diagnostics previously listed here were removed on 2026-10-21 — these sample-based measurement runs and the open-source tools remain free. This page does not quote or take payment. Enterprise buyer review has a separate, individually agreed scope; see buyer review. 中文入口：中文 FAQ（账单核对与免费任务） · README 中文版

What you get, what we ask

You get — the checkA measurement check on your sample: what safely counts as attempts vs operations, where retries may inflate usage against your billable unit, and what your telemetry cannot prove. Findings are stated as checkable claims; where your surface is blind, the report says UNPROVABLE instead of guessing.

You get — the resultA rerunnable result and, if you allow publication, your fixture committed under your handle with the claim boundary you state (synthetic or production — your call, carried verbatim).

We ask — never requiredPermission to publish the anonymized fixture (nothing customer-identifying; hash or drop identifiers). And if the check was useful, a star on the repo helps others find it — never required, never a condition.

If the check was useful, a star on the repo helps others find it — never required, never a condition.

How to start

Option 1 — GitHub issue #6 (preferred)

Comment on issue #6 — external measurement runs with your project/repo, protocol (MCP / API / other), and one agent-facing capability you want measured. The thread is public on purpose: the next provider reads your run's answer. First response within 24 hours.

Option 2 — email

Write to tongroy18@gmail.com with the same three lines and your sample attached or linked. Same terms: free, 24h first response, 48h check after a clean sample.

The full guide is docs/provider-onboarding.md — three entry paths, five steps, validate-before-sending, and the four criteria that define a successful run (the same yardstick for both sides).

Want to run the same review yourself? The full method — intake checks, the FMT-002 mapping guide, the mapper script and both report templates — is open: docs/review-method/.

Sample format: canonical JSONL per the collector input contract. python3 schemas/validate_jsonl.py your.jsonl must exit 0 before sending — that is the only format gate. Mapping method and the honest-absence discipline are in registry/provider-mapping.md.

Who has run one

External runs so far: a local agent runtime (availability certification vector), a metering provider (production receipt vectors with independent recomputation), and an x402 payment-surface provider (mapping of record agreed, fixture in progress). Fixtures live under conformance/ with their authors credited. More slots are open — the trial is expanding.

Prepare a summary locally first (optional)

verovo check --json report.json verovo share report.json --out summary.json verovo validate summary.json

--out writes immediately on your machine. Inspect the file before you choose to send it. Nothing is automatically uploaded. "No finding" and "not enough evidence" are valid outcomes.

What this is not

A free run is not an enterprise invoice review, a vendor acceptance, a refund, or runtime budget enforcement — those are separate tasks outside this scope. Tool bugs go to Issues free, with a redacted reproduction. Buyer-scope questions start at first-look.html.

The local check-up, the browser recount, the claims index and the conformance pack remain free and open source.