Connect an assistant

An MCP client that supports the MCP authorisation specification is intended to connect to an Autonym server, hosted or local. The server publishes its metadata at /.well-known/autonym and its MCP endpoint at /mcp. Which clients work in practice depends on the client and its connection options; tested combinations will be published as they are confirmed, and none is confirmed yet.

A remote connector and a locally configured server are different connection paths with different setup. A local address is not a universal instruction: a client that connects from its vendor's own infrastructure cannot reach a server on your machine.

Claude Desktop

Local mode runs over stdio. A hosted server is reached over Streamable HTTP at its /mcp endpoint, which Claude Desktop connects to as a remote server.

Add the local server to claude_desktop_config.json:

{
  "mcpServers": {
    "autonym": {
      "command": "npx",
      "args": ["-y", "autonym-local", "mcp"]
    }
  }
}

What the assistant sees

After the grant, the assistant holds six tools and one prompt. autonym_briefing is intended to return a one-paragraph summary of agent at L1 and communication at L1 for use in a system prompt, so the context is available to the assistant from the start of a conversation. What the assistant does with it is the assistant's decision, so responses can still vary.

Proposing entries

When the person tells the assistant something worth keeping, the assistant calls propose_entry with the domain, key, value and the person's words verbatim. The proposal waits in the review queue until the person confirms it. Where the client supports MCP elicitation, confirmation happens in the client's own interface, not by the assistant's say-so.