MARK

MARK is the Machine Agent Readiness Kit: a deterministic audit tool for product teams that want their websites, docs, and APIs to be usable by AI agents.

It checks for the signals an agent needs before it can safely reason about or call a product:

• Discovery: robots.txt, sitemap.xml, llms.txt, AGENTS.md, link hints
• Agent-readable content: structured llms.txt, markdown-friendly docs signals, JSON-LD
• Capability metadata: OpenAPI, MCP, API catalogs, plugin manifests
• Trust and auth: HTTPS, OAuth metadata, protected resource metadata, explicit crawler policy

MARK does not claim official conformance to AgentReady, MCP, or llms.txt. It is a private, CI-friendly readiness audit and remediation checklist.

For capability metadata, MARK checks stable well-known/root URLs first, then homepage/docs link hints. /llms.txt is useful backup discovery, but it should not be the only path to OpenAPI or MCP-style metadata for action-oriented products.

Run an audit

From npm without installing:

npx mark-agent-readiness-kit https://example.com --profile api

Or install the CLI:

npm install -g mark-agent-readiness-kit
mark-audit https://example.com --profile api

From a cloned repo:

npm run audit -- https://example.com --profile api

Write JSON:

npm run audit -- docs.example.com --profile docs --json --out report.json

Use a CI threshold:

npm run audit -- https://example.com --fail-under 70

Use a committed config file:

npm run audit -- --config mark.config.json

See docs/ci-quickstart.md for config keys and GitHub Actions usage.

Profiles:

• api: default; weights OpenAPI, MCP, and auth metadata heavily
• docs: weights discovery and agent-readable docs more heavily
• content: weights discovery and content structure over action/API metadata

Scoring details: docs/scoring.md

View the report UI

Open site/index.html in a browser and paste the JSON output from the CLI. The page also includes a built-in demo report.

Development

npm test
npm run check

Privacy model: docs/privacy.md