Connect over MCP

Mnemom operates two public Model Context Protocol servers. Both use the streamable-HTTP transport, so any MCP client that supports remote HTTP servers can connect.

Control plane

https://api.mnemom.ai/mcp62 tools mirroring the mnemom CLI — Trust Ratings, Alignment & Protection Cards, governance signals, postures, teams, webhooks, and API keys. tools/list is public; tool execution authenticates exactly as the REST API.

Docs search

https://docs.mnemom.ai/mcpRead-only. Two tools — search_mnemom_docs and query_docs_filesystem — over this documentation and the OpenAPI spec. No authentication.

The control plane authenticates with the same credentials as the REST API: a Bearer JWT or an X-Mnemom-Api-Key. See Authentication. The docs-search server needs no auth.

Authentication

Set your token once in the environment, then reference it from each client config:

export MNEMOM_TOKEN="<your Bearer JWT or mnm_… API key>"

The control-plane server passes the value through as Authorization: Bearer <token>. To use an API key instead, send it as the X-Mnemom-Api-Key header.

Client setup

Claude Code
Cursor
VS Code
Claude Desktop

Add both servers from the CLI:

# Control plane (authenticated)
claude mcp add --transport http mnemom https://api.mnemom.ai/mcp \
  --header "Authorization: Bearer $MNEMOM_TOKEN"

# Docs search (read-only, no auth)
claude mcp add --transport http mnemom-docs https://docs.mnemom.ai/mcp

List the registered servers with claude mcp list.

Add to ~/.cursor/mcp.json (global) or .cursor/mcp.json (per-project):

{
  "mcpServers": {
    "mnemom": {
      "url": "https://api.mnemom.ai/mcp",
      "headers": { "Authorization": "Bearer ${MNEMOM_TOKEN}" }
    },
    "mnemom-docs": {
      "url": "https://docs.mnemom.ai/mcp"
    }
  }
}

Add to .vscode/mcp.json in your workspace:

{
  "servers": {
    "mnemom": {
      "type": "http",
      "url": "https://api.mnemom.ai/mcp",
      "headers": { "Authorization": "Bearer ${MNEMOM_TOKEN}" }
    },
    "mnemom-docs": {
      "type": "http",
      "url": "https://docs.mnemom.ai/mcp"
    }
  }
}

Claude Desktop launches MCP servers as local processes, so bridge the remote endpoints with mcp-remote. Add to claude_desktop_config.json:

{
  "mcpServers": {
    "mnemom": {
      "command": "npx",
      "args": [
        "-y", "mcp-remote", "https://api.mnemom.ai/mcp",
        "--header", "Authorization: Bearer ${MNEMOM_TOKEN}"
      ]
    },
    "mnemom-docs": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://docs.mnemom.ai/mcp"]
    }
  }
}

Restart Claude Desktop after editing the config.

Verify the connection

tools/list is public on both servers, so you can confirm reachability with a raw JSON-RPC call before wiring up a client:

curl -sS -X POST https://api.mnemom.ai/mcp \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}' \
  | jq '.result.tools | length'

You should see the tool count (62 for the control plane, 2 for docs search). Tool execution on the control plane enforces the same authorization as the equivalent REST call — an unauthenticated client can enumerate tools but not invoke privileged ones.

Discovery

The control-plane server publishes a discoverable MCP server card (server.json shape) at https://www.mnemom.ai/.well-known/mcp/server-card.json. Clients that resolve server cards can configure themselves from that single URL.

Agent surface map

The full set of Mnemom’s agent-facing surfaces — discovery files, well-known metadata, and both MCP endpoints — is catalogued on the For Agents page.

​Connect over MCP

Control plane

Docs search

​Authentication

​Client setup

​Verify the connection

​Discovery

Agent surface map

Connect over MCP

Authentication

Client setup

Verify the connection

Discovery