Submit your tool for scoring

How review works

1. Request intake. Email us the basics: tool name, vendor, category, docs URL, and an OpenAPI spec or MCP endpoint if you publish one.
2. Provision credentials. A dedicated sandbox or demo tenant with a least-privilege API token created for us - never a shared production admin key. You keep revocation control at all times.
3. Supply test cases. At least three executable test cases in our submission format: the goal, the steps, and the expected result.
4. We run the review. Token-driven agents execute your test cases and our rubric line items against the environment you provisioned.
5. Score publishes. You get the full dimension breakdown and a 14-day right-of-reply window before the report goes live in the toolbox.

Credential requirements

• Sandbox or demo environment only - we never review against live customer data.
• Least-privilege token scoped to the surfaces your test cases exercise.
• Revocable by you at any time; we request a 30-day validity window.
• If you ship OAuth client-credentials, a service account, or an MCP server, send that - it scores better than a shared API key for a reason.

Credentials are stored encrypted, used only for review, and deleted when the report publishes.

Test-case format

Machine-readable, so our review agents can execute your cases directly. Schema: /api/submission-schema.json

{
  "id": "tc-01",
  "title": "Create an invoice via API",
  "goal": "Agent creates a draft invoice from a natural-language request",
  "steps": [
    "POST /v1/invoices with the sandbox customer id",
    "Attach one line item from the price book"
  ],
  "expected_result": "Draft invoice visible in the sandbox UI within 5s",
  "agent_accessible": true,
  "cleanup": "DELETE /v1/invoices/{id}"
}

Fair-review terms

• Scores cannot be purchased, sponsored, or negotiated. Ever.
• We publish what we find, including a low score. The right-of-reply window is for factual corrections, not tone.
• Re-review after material product changes is free - one per quarter.