Agent / MCP Surface Checker
One of the free tools from KinetixSEO, an SEO and AI-citation (GEO) checker. AI agents and MCP (Model Context Protocol) clients need a machine-readable way to discover what a site lets them do and how to connect — a different question from whether a search crawler can index the page. This tool probes for the surfaces that answer it: an MCP server descriptor, an A2A agent card, an OpenAPI specification, OAuth/OIDC discovery documents, agents.txt, auth.md, and several other emerging agent-native conventions, plus static proxies for whether the page's own content is actually readable and stable enough for an agent to act on.
Most sites score low here today simply because these conventions are new — a low score isn't a bug, it's an opportunity. Each missing item below links directly to what publishing it involves.
Frequently asked questions
What are MCP and A2A, and why does this tool check for them?
MCP (Model Context Protocol) is the emerging standard AI assistants and agent frameworks use to discover and call tools a site exposes — an "API menu" an agent can read without a human writing custom integration code. A2A (Agent-to-Agent) is a related standard for one autonomous agent to advertise its capabilities to another. A site that publishes an MCP server descriptor (/.well-known/mcp.json) or an A2A agent card (/.well-known/agent.json) is telling agents "here is exactly what you can do here, and how" — the same role robots.txt and sitemap.xml play for search crawlers, but for programmatic action rather than indexing.
My site is crawlable and indexed by Google — isn't that enough for AI agents too?
No — crawlability and agent-accessibility are different questions. A crawler reads your HTML to build a search index; an AI agent or MCP client is trying to discover callable capabilities (search this catalog, book this slot, place this order) and authentication/authorization rules it can act on programmatically. A page can be perfectly indexable and still have zero machine-readable surfaces telling an agent what it is allowed to do or how to authenticate — which is exactly what this checker measures, distinct from the SEO/GEO citability checks in our full audit.
What do OpenAPI, OAuth discovery, and agents.txt each do?
An OpenAPI specification (/openapi.json or /openapi.yaml) is a machine-readable contract describing your API's endpoints, parameters, and responses, so an agent can call it directly instead of scraping HTML. OAuth 2.0 Authorization Server Metadata (/.well-known/oauth-authorization-server) and OpenID Connect discovery (/.well-known/openid-configuration) let an agent find your authentication endpoints and supported grant types automatically rather than having a developer hard-code them. agents.txt is a simpler, newer plain-text convention (alongside llms.txt) stating in prose what actions AI agents may take on your site and at what rate — useful even before you have a real API for agents to call.
How do I actually add these surfaces to my site?
Each missing item in the checklist below links its exact fix, but the short version: publish static files at the conventional well-known paths (an MCP descriptor at /.well-known/mcp.json, an A2A card at /.well-known/agent.json, an OpenAPI document at /openapi.json, agents.txt at your site root) — most of these are static JSON or text files a build step can generate once and re-deploy, not a running service. If you don't control application code directly, a reverse proxy or CDN edge rule that serves a static file at each well-known path is enough to pass these checks.
Why do accessibility landmarks and stable layout affect an "agent surface" score?
An agent that navigates a page via its accessibility tree (rather than a full visual render) relies on the same landmarks — <main>, <nav>, a valid heading hierarchy, alt text — a screen reader does; missing them makes the page structurally unreadable to that class of agent, not just to assistive technology. Layout stability matters for a related reason: an agent that reads or clicks at a specific coordinate mid-navigation can hit the wrong element if content reflows under it a moment later, the same failure mode Cumulative Layout Shift causes for a human.
This is one check from KinetixSEO's full SEO/GEO audit. Want the complete picture — Core Web Vitals, AI-citation readiness, technical SEO health, and tailored fixes? Run the free full checker.