Capabilities — Exposing Tools an Agent Can Call

The agent-readiness dimension that exposes tools an agent can call — a JSON API, MCP server, WebMCP and agents.json — so agents act, not just read.

What the capabilities dimension means

The capabilities dimension turns your site from something an agent reads into something an agent can use. Discovery, content and access control get an agent to your information; capabilities give it typed actions it can invoke — search a catalog, check a booking, submit a form, run a query. You only need this dimension if agents should do something on your behalf rather than just consume facts, so it sits later in the implementation order. When you do expose tools, the goal is a declared, machine-callable interface rather than asking the agent to drive your human UI.

Signals and standards it covers

How the Agent-Readiness Audit scores it

The Audit scores capabilities on whether a declared, callable action surface exists and resolves. The anchor check is capabilities.agents_json: it passes when an agents.json manifest is present, parseable, and declares at least one well-formed action. Companion checks confirm that a referenced JSON API or MCP/WebMCP endpoint is reachable and responds in the expected shape. The criterion is concrete — the manifest parses and the endpoint answers, or it does not. Verify the exact agents.json schema against its primary spec at build, since the action format evolves.

Related: the agents.json spec · publish agents.json · the Agent Protocol Atlas · frontier models that complete tasks · audit your site

← Agent-Readiness · .md