Google Colab MCP Bridge

A persistent web application and agent execution bridge for Google Colab Model Context Protocol (MCP).

Why this exists

Normally, executing code through standard MCP scripts starts a new server process every run, causing Google Colab to open a brand-new browser tab on every single command.

gcolab_mcp_web_mcp solves this:

• It runs a single persistent background daemon that pairs with your Google Colab tab once.

• You can execute unlimited commands, scripts, machine learning workloads, or shell commands with zero new tabs opened.

• Exposes both a sleek dark-mode Web Dashboard and a REST/JSON API for AI agents.

Related MCP server: colab-mcp

Quick Start

1. Launch the Application

You can now start it with any of these simple commands (no flags needed):

./start.sh

or

python3 run.py

or

uv run run.py

(Dependencies are declared inside run.py and are automatically handled on the fly).

2. Open the Dashboard

Navigate to http://localhost:8000 in your browser.

3. Connect Colab Once

• Click "Connect Colab Tab" (or open the provided URL in your browser).

• Your Colab tab will display: "Connected to local Colab MCP server".

• The dashboard status indicator turns 🟢 Connected to Colab.

Features

1. One-Click Cloud Code Execution:

   • Write Python or shell commands (e.g., !nvidia-smi, !pip install) and run them directly on Colab's cloud GPU.

   • Shortcut: Ctrl + Enter to execute.

2. Real-time Terminal Output:

   • Live stdout, stderr, execution timer, copy output, and plot support.

3. Notebook Cell Manager:

   • Inspect all existing cells in your Google Colab notebook, run specific cells, or delete cells directly from the dashboard.

4. W3C WebMCP Standard Support (Browser Native):

   • Exposes browser tools via navigator.modelContext.registerTool:

     • execute_python_on_colab: Runs code remotely on the Colab kernel.

     • get_colab_notebook_cells: Fetches notebook cells.

     • check_colab_status: Checks connection state.

   • Annotated Declarative Form (toolname="executeColabCode").

   • Compliant HTTP headers (Permissions-Policy: tools=(self), Origin-Agent-Cluster: ?1).

   • Compatible with Chrome Canary (chrome://flags/#enable-webmcp-testing) and includes an automatic in-page polyfill and tester.

5. AI Agent Integration API (REST):

   • External agents (or scripts) can send code to Colab via a simple HTTP request:

curl -X POST http://localhost:8000/api/execute \
  -H "Content-Type: application/json" \
  -d '{"code": "import torch; print(torch.cuda.is_available())"}'

import requests

response = requests.post("http://localhost:8000/api/execute", json={
    "code": "!nvidia-smi"
})
print(response.json()["output"])