Colab MCP Server

Control and interact with Google Colab instances via a reverse proxy

README.md

Colab MCP Client (TypeScript)

A high-performance Model Context Protocol (MCP) server for controlling Google Colab instances via a reverse proxy (ngrok, serveo, etc.).

Features

  • connect_to_colab: Link to your running Colab instance.
  • run_colab_shell: Execute shell commands.
  • run_colab_python: Run persistent Python code.
  • get_colab_system_info: Monitor CPU/GPU/RAM.

Quick Start (No Install)

You can run this MCP server directly using Bun:

bun https://github.com/DevAdalat/colab-mcp/releases/download/v1.0.0/colab-bridge.js

Adding to MCP Agents (Claude Desktop, etc.)

For a stable setup, it is recommended to download the file locally:

  1. Download the binary/script:

    curl -LO https://github.com/DevAdalat/colab-mcp/releases/download/v1.0.0/colab-bridge.js
    
  2. Add to Claude Desktop: Update your claude_desktop_config.json (usually at ~/Library/Application Support/Claude/claude_desktop_config.json on macOS):

{
  "mcpServers": {
    "colab-bridge": {
      "command": "bun",
      "args": [
        "/absolute/path/to/colab-bridge.js"
      ]
    }
  }
}

Manual Installation

If you want to run it locally:

  1. Install Dependencies:

    bun install
    
  2. Run Server:

    bun run index.ts
    

Requirements

  • Bun runtime installed.
  • A running Google Colab instance with the bridge script active.

Tools 4

connect_to_colabLink to your running Colab instance.
run_colab_shellExecute shell commands on the connected Colab instance.
run_colab_pythonRun persistent Python code on the Colab instance.
get_colab_system_infoMonitor CPU, GPU, and RAM usage of the Colab instance.

Try it

Connect to my running Colab instance at the provided URL.
Run a shell command to list the files in the current Colab directory.
Execute this Python script to train my model on the Colab instance.
Check the current GPU and RAM usage on my Colab instance.

Frequently Asked Questions

What are the key features of Colab MCP?

Connect to remote Google Colab instances. Execute arbitrary shell commands remotely. Run persistent Python code blocks. Monitor real-time system resource usage (CPU, GPU, RAM).

What can I use Colab MCP for?

Running long-term data processing tasks on Colab while interacting via Claude. Monitoring GPU availability and usage during model training. Executing shell-based environment setup scripts on remote notebooks. Managing file systems on Colab instances through natural language commands.

How do I install Colab MCP?

Install Colab MCP by running: curl -LO https://github.com/DevAdalat/colab-mcp/releases/download/v1.0.0/colab-bridge.js

What MCP clients work with Colab MCP?

Colab MCP works with any MCP-compatible client including Claude Desktop, Claude Code, Cursor, and other editors with MCP support.

Turn this server into reusable context

Keep Colab MCP docs, env vars, and workflow notes in Conare so your agent carries them across sessions.

Open Conare