Skip to main content
Every Fluent agent exposes its own Model Context Protocol server over Streamable HTTP. An MCP client that connects to it gets the tools that agent is configured with — the same tools the agent uses itself.
The transport is stateless: each request carries everything the server needs, and there is no session to keep open. DELETE on the endpoint is accepted and acknowledged so well-behaved clients can run their cleanup step.

​
Authenticating

Both credential types work:
An agent-scoped OAuth token is the better fit for a client acting on a person’s behalf. A request with no credential receives a 401 with an RFC 9728 WWW-Authenticate challenge naming the protected-resource metadata document, so a spec-compliant MCP client can discover the authorization server and complete the OAuth flow on its own:
The authorization server itself is described at:

​
Configuring a client

Most MCP clients accept a remote server as a URL plus a header. For example, in a client that reads a JSON config:
List what an agent offers with a plain JSON-RPC call:
Use the API index to discover the current endpoints, scopes, and rate limits without authenticating.

​
Limits

MCP requests count against the same rate limit as the rest of the API: 100 requests per 60 seconds per credential. See Errors and rate limits.