Setup guide

Connecting OmniMem to Cursor

Cursor is the AI code editor built on VS Code. It supports Streamable HTTP MCP servers, so it connects straight to OmniMem's /mcp endpoint.

Quick setup

Create or edit ~/.cursor/mcp.json for global access:

{
  "mcpServers": {
    "omnimem": {
      "url": "http://127.0.0.1:8765/mcp"
    }
  }
}

Using the desktop app? Copy MCP URL in the tray or menu bar menu gives you the address.

If OmniMem has an access token set (MCP_AUTH_TOKEN, or Access token on the Configuration page), send it as a header:

{
  "mcpServers": {
    "omnimem": {
      "url": "http://127.0.0.1:8765/mcp",
      "headers": {
        "Authorization": "Bearer your-token-here"
      }
    }
  }
}

Or keep the token out of the file with environment variable interpolation:

{
  "mcpServers": {
    "omnimem": {
      "url": "http://127.0.0.1:8765/mcp",
      "headers": {
        "Authorization": "Bearer ${env:OMNIMEM_TOKEN}"
      }
    }
  }
}

Cursor uses ${env:VAR_NAME}, not ${VAR_NAME}.

Coming from 6.x, just change /sse to /mcp.

Global vs project config

File Scope Use when
~/.cursor/mcp.json All projects OmniMem should be available everywhere (recommended)
.cursor/mcp.json (project root) Single project Only this project needs OmniMem

Project config wins when the same server name is in both. Some Cursor versions have had a bug where the global file is silently ignored, so if the tools don't appear, try the project file.

Enabling MCP tools

MCP tools only show up in Cursor's Agent mode, not plain chat. After adding the config:

  1. Restart Cursor or reload the window
  2. Open Settings (Cmd/Ctrl + ,) and search for MCP
  3. Check omnimem is listed and enabled
  4. Switch the chat panel to Agent mode

Cursor asks before each MCP tool call by default. There's a separate setting to auto-run MCP tools if the prompts drive you round the bend.

Using OmniMem in Cursor

Once it's connected, just ask:

Known quirks

Verifying the connection

In Agent chat, ask:

Can you call the OmniMem health tool?

You should get back the record and vector counts per namespace, whether the embedding model is loaded, and the uptime.

Notes

By Ric HarveyUpdated All guidesThis guide on Forgejo