What it does

A deterministic, local-first CLI compares the claims in an MCP, API, or SDK listing against captured manifest metadata and returns a tri-state result: pass, fail, or unknown. Unmeasured is never reported as fail. Reports export hashed claim and evidence references and reasons only — never raw listing text, prices, or secrets.

Read the methodology for how a result is decided, or the glossary for the exact meaning of each term.

Honesty rules

Deterministic and versioned: the same input always yields the same report.
A missing capability stays unknown unless the capture explicitly asserts completeness.
No live fetch, no automatic edits to any third-party listing.
Hosted evaluation, history, and billing are disabled; the CLI runs on your machine.

Supported adapters

mcp-tools-list-v1 — MCP manifest tools to capability metadata.
openapi-operations-v1 — captured OpenAPI operations and pricing plans to capability metadata.

Install and run

Clone the repository and run the CLI against a local fixture:

git clone https://github.com/gitdhillonai/listingtruth.io
cd listingtruth.io
node src/cli.mjs fixtures/failed.json

Print the versioned input/output JSON Schema contract with node src/cli.mjs --schema.