Check a domain agentic web readiness
Enter a domain. We fetch its public agent-facing paths (about 30, no credentials), validate every manifest we find against its specification, and grade the result. The report has a shareable URL and a badge.
How the grade is made
+10 per lifecycle step with a standard observed, counting manifests that validate or that the probe rule for the standard recognises (a valid MCP server card counts for MCP; max 90); +3 per observed standard beyond the first two (max 15); conformance +15 (+10 with one manifest) minus 5 per agent-facing manifest with errors, never below 0 (security.txt is reported, not scored). A+ >= 60, A >= 45, B >= 30, C >= 20, D >= 10.
- Errors are specification violations: a missing required field, the wrong media type, invalid JSON.
- Warnings are best-practice gaps: no caching headers on a server card, no summary line in
llms.txt, an apex that redirects towww. - Standards that need credentials or have no fixed path are
skipped, not counted either way.
Manifests validated
| Standard | Path | What is checked |
|---|---|---|
| Agent Resource Discovery (ARD) | /.well-known/ard.json | JSON object; specVersion; entries[] each with identifier, type and url or data; urn:air: identifiers; host block |
| AI Catalog | /.well-known/ai-catalog.json | same checks as ARD |
| MCP Server Cards (SEP-2127) | /mcp/server-card, /.well-known/mcp.json | $schema v1, reverse-DNS name, version, description ≤ 100, remotes[] with streamable-http or sse and a URL; media type, CORS, caching; no primitives in the card |
| A2A agent card | /.well-known/agent-card.json | name, url, version, skills[] with id, name, description; capabilities, default modes, protocolVersion, provider |
| llms.txt | /llms.txt | H1 title first; blockquote summary; Markdown links; text media type; size |
| AGENTS.md | /AGENTS.md | non-empty Markdown with an H1 |
| API catalog (RFC 9727) | /.well-known/api-catalog | application/linkset+json; linkset[] with anchor and a service-desc, service-doc or service-meta link; href present, type on service-desc |
| AI crawler tokens in robots.txt | /robots.txt | User-agent groups; named AI tokens; Sitemap; optional Agentmap and Content-Signal |
| security.txt (RFC 9116) | /.well-known/security.txt | Contact and a future Expires required; Canonical recommended; text/plain |
| auth.md | /auth.md | Markdown naming at least one mechanism (identity_assertion, service_auth, anonymous) |
Recently checked
- 2s.io
- agenticresourcediscovery.org
- airtable.com
- amplitude.com
- anthropic.com
- api.mcp.github.com
- apify.com
- aws.amazon.com
- bird.com
- cisco.com
- clickhouse.com
- cloudflare.com
For agents
The same report is the validate_manifest tool on the MCP server: validate_manifest({"origin": "example.com"}), or one manifest with id. Shareable JSON at /check/<domain>/index.json, Markdown at /check/<domain>/index.md, badge at /badge/<domain>.svg.
