Connecting MemPalace to Cursor, GitHub Copilot, and Gemini CLI

Cursor IDE

Cursor has native MCP support. Once configured, the AI inside Cursor can call all 19 MemPalace tools automatically when it determines they are relevant to your question — no manual command invocation needed.

Configuration

Create or edit the file ~/.cursor/mcp.json:

{
  "mcpServers": {
    "mempalace": {
      "command": "python",
      "args": ["-m", "mempalace.mcp_server"]
    }
  }
}

Restart Cursor after saving this file. Cursor will launch the MCP server as a child process the next time you open the AI panel.

Verifying Cursor

Open a Cursor AI chat session and type: "Can you check my MemPalace status?" If the connection is working, Cursor will call mempalace_status and display your wing counts, the AAAK spec, and the memory protocol. You should see your actual palace data — not a generic response.

GitHub Copilot in VS Code

Copilot Chat in VS Code 1.x+ supports MCP tools through the built-in MCP configuration. The server config lives in your workspace rather than a global settings file, which means each project can have its own MCP configuration.

Configuration

Create .vscode/mcp.json in your project root:

{
  "servers": {
    "mempalace": {
      "type": "stdio",
      "command": "python",
      "args": ["-m", "mempalace.mcp_server"]
    }
  }
}

VS Code will detect the file and prompt you to approve the new MCP server. Once approved, Copilot Chat can invoke MemPalace tools using the @mempalace mention: @mempalace search "auth migration decisions".

Making It Global

To avoid adding the file to every project, add the same configuration to your user-level VS Code settings under mcp.servers. The exact path depends on your operating system: ~/Library/Application Support/Code/User/settings.json on macOS, ~/.config/Code/User/settings.json on Linux.

Gemini CLI

Gemini CLI has the deepest MemPalace integration of any currently supported tool — it auto-configures the MCP server connection, the session save hook, and the wake-up context loading without requiring manual setup of each component.

Installation

pip install mempalace
pip install gemini-cli
# Follow the setup guide at:
# github.com/milla-jovovich/mempalace/blob/main/examples/gemini_cli_setup.md
gemini
# MemPalace context loads automatically at session start

After the one-time guided setup, Gemini CLI handles three things automatically on every session: loading L0+L1 context at start, connecting the MCP server so Gemini can call all 19 tools, and triggering a save at the end of the session so the palace is kept current.

Testing Any MCP Connection

Regardless of which tool you are using, the same three prompts will confirm that the connection is working and MemPalace has your actual data:

• "What does mempalace_status return?" — should display your actual wing count, not an error or a generic response
• "Search MemPalace for [a topic you know is in your conversations]" — should return a recognisable verbatim passage
• "List my MemPalace wings" — should list the wings created during your mempalace init setup

If any of these fail with a tool-not-found error, the MCP server is not connected. If they return empty results, the server is connected but your palace may not have been mined yet.

Saving Sessions Automatically

Cursor and Copilot do not have a built-in hook mechanism like Claude Code does. The simplest approach is a shell alias that triggers a manual save:

# Add to your ~/.zshrc or ~/.bashrc
alias mpsave="bash ~/.mempalace/hooks/mempal_save_hook.sh"
# Run at the end of each significant coding session
mpsave

For a more automated approach on macOS, the save hook can be added to a launchd scheduled job that runs every hour. On Linux, the equivalent is a cron entry. Either approach ensures your palace stays current without requiring a manual trigger after every session.