Skip to main content
Documentation
Docs / Products / Checks
Agent setup

Verify

Checks

Compare software claims with observed checks and review what is verified, blocked, failed, or still untested.

Last updated August 22, 2026

See what the evidence supports before you ship.

Use cases

Review the claim and its basis

See what was expected and what was actually observed.

Keep untested work visible

A reachable page does not turn an untested signup or payment flow into a pass.

Give the next test a purpose

Use unresolved claims to decide what needs a test, a repair, or a human decision.

How it works

  1. 1

    Describe the behavior

    Enter the website URL and the claims you want to check.

  2. 2

    Collect and supply evidence

    Run the HTTP observation and add results from the relevant workflow tests.

  3. 3

    Review the result

    Inspect findings and resolve the remaining claims before approving release.

Quickstart

Open Checks in the console, provide the required inputs, and start an organization scoped run.

Inputs

  • GitHub repository, branch, and commit
  • deployed URL and environment label
  • approved product claims
  • test identities and roles
  • scan policy and budget

Outputs

  • completion graph with denominator
  • evidence bundle per finding
  • typed repair packets
  • release gate verdict
  • truth diff between runs
Open Checks

MCP

Call Checks from a compatible agent with the preferred public tool name.

bashReady
symbolic.release.run

The previous internal service identifier remains accepted for compatibility, but new integrations should use this product name.