# Capabilities

Can an agent do something here, not just read?

Tools, endpoints and contracts an agent can call: the newest and fastest-moving layer, where the gap between a protocol's momentum and the surface a crawl can actually see is widest. A capability advertised and unreachable is worse for an agent than none advertised at all.

| Technology | What it is | Adoption | Status |
|---|---|---|---|
| [WebMCP](/docs/capabilities/web-mcp) | In-page tools a browsing agent can call directly. | Early pilots, ▲ rising | scored |
| [MCP server card](/docs/capabilities/mcp-server) | Advertises a Model Context Protocol endpoint agents can call. | Early pilots, ▲ rising | scored |
| [MCP authorization](/docs/capabilities/mcp-auth) | The OAuth discovery chain protecting that MCP endpoint. | Early production, ▲ rising | scored |
| [OpenAPI description](/docs/capabilities/open-api) | A machine-readable contract for the site's API. | Web baseline, ► flat | scored |
| [NLWeb endpoint](/docs/capabilities/nl-web) | Microsoft's conversational /ask route returning schema.org answers. | Early pilots, ► flat | scored |
| [schema:Action](/docs/capabilities/schema-action) | Declares actions an agent could take from this page. | Web baseline, ► flat | scored |
| [auth.md](/docs/capabilities/auth-md) | Telling an agent how to register for an account and get its own credentials, without a signup form or a consent screen. | Early pilots, ▲ rising | watched |