# WebMCP and agent discovery checker

Check the protocols AI agents use to discover what a site can do: WebMCP tools in the page, and the MCP, A2A, agent skills, API catalog and OAuth documents under /.well-known.

Run it at https://crawlwise.dev/tools/mcp-agent-discovery-checker: paste a public URL, no account needed. The scan runs the full CrawlWise audit and returns a private report.

## What this checker tests

- WebMCP tools are complete (`api.webmcp`)
- An MCP server card describes the site's MCP server (`api.mcp-server-card`)
- An A2A agent card describes the site's agent (`api.a2a-agent-card`)
- An agent skills index lists the site's skills (`api.agent-skills`)
- An API catalog lists the site's APIs (`api.catalog`)
- OAuth or OpenID Connect discovery is published (`api.oauth-discovery`)
- OAuth protected resource metadata is published (`api.oauth-protected-resource`)
- An ARD manifest lists the site's agent resources (`api.ard`)

## What the checker reads

The scan records every tool the page registers through WebMCP, with its name, description and
input schema, and fetches each discovery document from `/.well-known`:

| Document | Path |
|---|---|
| MCP server card | `/.well-known/mcp/server-card.json` |
| A2A agent card | `/.well-known/agent-card.json` |
| Agent skills index | `/.well-known/agent-skills/index.json` |
| API catalog | `/.well-known/api-catalog` |
| OAuth / OpenID discovery | `/.well-known/oauth-authorization-server`, `/.well-known/openid-configuration` |
| OAuth protected resource | `/.well-known/oauth-protected-resource` |
| ARD manifest | `/.well-known/ard.json` |

These protocols are new and several are drafts, so the checks are experimental with weight 0:
they appear in their own report tab and never change the score.

## Start with the basics

Agents still have to reach and read your pages first. Run the
[AI crawler checker](/tools/ai-crawler-checker) and the
[JavaScript test](/tools/ai-javascript-checker) before adding protocols, and read the
[agent-ready checklist](/guides/agent-ready-website-checklist) to see which of them your kind of
site needs. The [WebMCP guide](/guides/webmcp) shows how to declare a tool.

## FAQ

### Does my website need an MCP server or WebMCP?

Only if agents should take actions on it or use its API. A content or marketing site is agent-ready without them. CrawlWise reports a missing document as not applicable and never lowers the score for it.

### What is WebMCP?

A proposed browser API that lets a page register tools, such as search or booking, that a browser agent can call with structured input instead of clicking through the interface. Chrome offers it in an origin trial.

### Where should an MCP server card live?

The draft puts it at /.well-known/mcp/server-card.json, with at least serverInfo.name. An A2A agent card lives at /.well-known/agent-card.json.

### Why does my single-page app fail these checks?

Many single-page apps answer every path with their HTML shell and status 200. An HTML page at a /.well-known path is not a discovery document, so CrawlWise reports it as missing.

## Related guides

- [WebMCP: let AI agents use your site's forms](https://crawlwise.dev/guides/webmcp.md)
- [Is your site agent-ready? A practical checklist](https://crawlwise.dev/guides/agent-ready-website-checklist.md)
- [Lighthouse Agentic Browsing: what it checks](https://crawlwise.dev/guides/lighthouse-agentic-browsing.md)

---

Last updated 14 November 2026.

HTML version: <https://crawlwise.dev/tools/mcp-agent-discovery-checker>

Every CrawlWise page in Markdown: <https://crawlwise.dev/llms.txt>
