# Coverage and receipts

A Check receipt records what ran, what it cost, and what was not engaged. Silence is never clearance.

Every governed run leaves a coverage register and, when the run is terminal, a receipt. The register is part of the result: checks that did not apply or could not run are visible. Not engaged is not billed as completed work, and it is not implied pass.

## Coverage register

- Engaged: the criterion ran against the deliverable.
- Not engaged: the criterion did not run. The receipt says why it was skipped or inapplicable.
- Undetermined: bounded work ran and the verifier abstained. Abstention still costs, and it is not a pass.

## What a receipt binds

A managed receipt binds the run identity, pack stack, admission, verifier digest, budget ceiling, and terminal. Proposal, estimate, and governed decision remain different output authorities. A worker callback is a wake signal; the host reads sealed artifacts before committing a terminal.

## Findings are locators

Public MCP and API reads return disposition summaries and locators, not document excerpts. Do not expect this website to host findings payloads. Authorized agents poll `get_findings` / `get_receipt` with the durable `checkId`.

> Keep private artifacts out of crawlers and prompt context. This site lists how Check works. It does not list customer documents, quotes, or receipts.

## Billing boundary (pilot)

The unit is governed work, not the verdict: a clean result still required execution. Public self-service rates are not approved; pilot pricing is operator-provisioned. See [pricing](/pricing) for the human-readable summary.