Skip to content

Setting Up Penfield with MCP-Remote

Penfield works with any AI platform that supports MCP (Model Context Protocol) via mcp-remote. This guide covers setup for popular platforms.


Prerequisites

Most platforms require Node.js to run mcp-remote. If you don't have Node.js installed, download it from nodejs.org.


Configuration Reference

Most platforms use the same core configuration. Copy and adapt as needed:

{
  "mcpServers": {
    "Penfield": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://mcp.penfield.app/"
      ]
    }
  }
}

Note: If you already have other MCP servers configured, add the "Penfield": { ... } block inside your existing "mcpServers" object. Don't create a second "mcpServers" key.


Authorization

The first time you use Penfield in any application, a browser window will open asking you to authorize access. After authorizing once, other applications will share the same mcp-remote authorization automatically. If mcp-remote is upgrade, a new authorization may be required.

If authorization doesn't trigger automatically, try restarting the application after adding the configuration.


Platform-Specific Instructions


Antigravity

Step 1: Open Agent Panel Menu

In the Agent panel on the right side of the screen, click the three-dot menu (⋯).

Click the three-dot menu in Agent panel

Step 2: Open MCP Servers

From the dropdown menu, click MCP Servers.

Click MCP Servers

Step 3: Open Manage MCP Servers

In the MCP Store panel, click Manage MCP Servers in the top right.

Click Manage MCP Servers

Step 4: View Raw Config

Click View raw config to open the configuration file.

Click View raw config

This opens ~/.gemini/antigravity/mcp_config.json for editing.

Step 5: Add Configuration

Add the Penfield configuration and save the file:

Antigravity mcp_config.json with Penfield

{
  "mcpServers": {
    "Penfield": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://mcp.penfield.app/"
      ]
    }
  }
}

Step 6: Restart and Test

Restart Antigravity. If you haven't authorized Penfield previously, a browser tab will open automatically. Start a new chat and type "Penfield Awaken" to test the connection.


Perplexity

Perplexity Desktop has a built-in connector interface that doesn't require editing JSON files directly.

Step 1: Open Settings

Click the Settings icon in Perplexity Desktop, then click Connectors.

Perplexity settings showing Connectors option

Step 2: Add Connector

Click Add Connector.

Connectors page with Add Connector option

Step 3: Configure (Simple View)

On the Simple tab, enter:

Field Value
Server Name Penfield
Command npx -y mcp-remote https://mcp.penfield.app/

Simple configuration view

Step 4: Verify Advanced Settings

Click the Advanced tab to confirm the JSON matches:

Advanced view showing JSON configuration

Step 5: Save and Restart

Click Save, then restart Perplexity Desktop. If you haven't authorized Penfield previously, a browser window will open for authorization.


Gemini CLI

Step 1: Edit Settings File

Open ~/.gemini/settings.json in your preferred editor:

nano ~/.gemini/settings.json
# or
code ~/.gemini/settings.json

Step 2: Add Penfield Configuration

Add the Penfield configuration to your settings file. Your file structure may vary depending on your authentication method:

Gemini CLI settings.json with Penfield configuration

{
  "mcpServers": {
    "Penfield": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://mcp.penfield.app/"
      ]
    }
  }
}

Step 3: Restart and Authorize

Run gemini again. If you haven't authorized Penfield previously, the authorization flow will open in your browser automatically. After authorizing, quit and restart Gemini CLI.


Windsurf

Step 1: Install mcp-remote

If you haven't already, install mcp-remote:

npm install -g mcp-remote

Step 2: Open MCP Configuration

Click the plug icon (1), then click the settings icon (2) to open the configuration file.

Windsurf showing MCP plugin and settings icons

This opens ~/.codeium/windsurf/mcp_config.json. If this is your first MCP server, the file may be empty.

Step 3: Add Configuration

Add the Penfield configuration:

Windsurf mcp_config.json with Penfield

{
  "mcpServers": {
    "Penfield": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://mcp.penfield.app/"
      ]
    }
  }
}

Step 4: Restart and Verify

Quit and restart Windsurf. If you haven't authorized Penfield previously, a browser tab will open automatically.

After authorization, click the plug icon to verify Penfield is connected with 18 tools available:

Windsurf showing Penfield connected with 18 tools


Cursor

Known Issue: Cursor's MCP authentication can be unreliable. If you experience problems, try authorizing Penfield in another application first (such as Windsurf, Gemini CLI, or LM Studio), then return to Cursor.

Note: MCP tools may only be available in Agent mode, not in Ask or Plan modes.

Step 1: Open MCP Settings

  1. Click the gear icon to open settings
  2. Click Tools & MCP
  3. Click Add Custom MCP

Cursor settings showing Tools & MCP

This opens ~/.cursor/mcp.json for editing.

Step 2: Add Configuration

Add the Penfield configuration:

Cursor mcp.json with Penfield configuration

{
  "mcpServers": {
    "Penfield": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://mcp.penfield.app/"
      ]
    }
  }
}

Step 3: Restart

Save the file and restart Cursor. If authorization doesn't trigger automatically, you may need to restart again after authorizing through another application.


LM Studio

Important: Local models have significant limitations compared to frontier models. See LM Studio Notes below.

Step 1: Open MCP Configuration

Click the Program tab, click the Install dropdown, and select Edit mcp.json.

LM Studio showing Edit mcp.json option

Step 2: Add Configuration

Add the Penfield configuration and save the file:

LM Studio mcp.json configuration

{
  "mcpServers": {
    "Penfield": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://mcp.penfield.app/"
      ]
    }
  }
}

Step 3: Restart and Load Model

Restart LM Studio and load your model. If you haven't authorized Penfield previously, a browser tab will open automatically.

LM Studio Notes

Model requirements:

  • Only models with explicit tool use support will work with Penfield
  • Set Context Length to the maximum your hardware supports—MCP tool calls generate long contexts that can overflow smaller windows

LM Studio model settings showing context length

Performance expectations:

Local models typically have single or low double-digit billions of parameters, while frontier models operate in the hundreds of billions or trillions. Expect correspondingly different output quality.

Recommended system prompt:

You are my personal AI assistant and you will use Penfield to manage recalling and storing memories.

Other Platforms

Any AI platform that supports mcp-remote can use Penfield. Add this configuration to your platform's MCP configuration file:

{
  "mcpServers": {
    "Penfield": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://mcp.penfield.app/"
      ]
    }
  }
}

Common configuration file locations:

Platform Configuration File
Antigravity ~/.gemini/antigravity/mcp_config.json
Windsurf ~/.codeium/windsurf/mcp_config.json
Cursor ~/.cursor/mcp.json
Gemini CLI ~/.gemini/settings.json
LM Studio Via Program > Install > Edit mcp.json

Troubleshooting

Authorization doesn't trigger: - Restart the application after adding the configuration - Try authorizing through a different application first

"Command not found" or npm errors: - Ensure Node.js is installed: node --version - Try installing mcp-remote globally: npm install -g mcp-remote

Tools not appearing: - Verify the JSON syntax is correct (no trailing commas, matching braces) - Check that you're in the correct mode (e.g., Agent mode in Cursor) - Restart the application

Still having issues? Contact [email protected] for assistance.