Enterprise DNA
M MCP Servers Developer low

SebastianGilPinzon/colab-mcp

by Various

Fixed & enhanced fork of Google's Colab MCP — tools visible at startup, GPU control, Windows support, no more 'Disconnected'

S

MCP

SebastianGilPinzon/colab-mcp

Added 24 June 2026

#claude-code #cursor #google-colab #gpu #llm-tools #mcp #model-context-protocol

Overview

A fixed and enhanced fork of Google's Colab MCP that makes tools visible at startup, adds GPU control, and supports Windows. It resolves the common 'Disconnected' error that plagues the original implementation.

Best for

Best for
Developers who need a reliable Colab MCP fork with GPU control and Windows compatibility

Use cases

  • Run AI model training scripts on Colab GPUs from a local MCP client
  • Manage Colab notebook execution with GPU toggling and startup tool visibility
  • Develop and test MCP integrations on Windows without disconnection issues

How to use

Tools exposed

  • change_runtime
  • open_colab_browser_connection
  • add_code_cell
  • add_text_cell
  • get_cells
  • run_code_cell
  • update_cell
  • delete_cell
  • move_cell

Tested with

Claude Code, Cursor, Windsurf, VS Code

Example client config

{\n  "mcpServers": {\n    "colab-proxy-mcp": {\n      "command": "uv",\n      "args": [\n        "run", "--directory", "/path/to/colab-mcp",\n        "colab-mcp",\n        "--client-oauth-config", "/path/to/colab-oauth.json"\n      ],\n      "timeout": 30000\n    }\n  }\n}

Notes

A fixed and enhanced fork of Google’s Colab MCP that makes tools visible at startup, adds GPU control, and supports Windows. It resolves the common ‘Disconnected’ error that plagues the original implementation.

6 stars on GitHub. Last updated 2026-06-19. Licensed Apache-2.0.

Use cases

  • Run AI model training scripts on Colab GPUs from a local MCP client
  • Manage Colab notebook execution with GPU toggling and startup tool visibility
  • Develop and test MCP integrations on Windows without disconnection issues

Pros

  • Fixes the persistent ‘Disconnected’ bug from the original Colab MCP
  • Adds GPU control and Windows support missing from the upstream fork
  • Tools are visible immediately at startup, improving developer workflow

Cons

  • Very small community with only 6 GitHub stars and limited documentation
  • Depends on Google Colab’s infrastructure which can change without notice
  • No active maintenance guarantees beyond the current fork state

Indexed from awesome-mcp-servers-punkpeye and enriched against its public facts.

Pros

  • Fixes the persistent 'Disconnected' bug from the original Colab MCP
  • Adds GPU control and Windows support missing from the upstream fork
  • Tools are visible immediately at startup, improving developer workflow

Cons

  • Very small community with only 6 GitHub stars and limited documentation
  • Depends on Google Colab's infrastructure which can change without notice
  • No active maintenance guarantees beyond the current fork state
Free 27-page guide

Get the free Developer’s Field Guide

A 27-page field guide to the AI coding workflow with Claude. Claude Code, MCP servers, the prompt patterns that work, and what to delegate. Free.

Enter your work email. We send it straight over, plus a few short notes worth knowing. Unsubscribe any time.

No spam. Unsubscribe any time.

Running a business, not writing the code? See the MCP servers picked for operators, and get your first one wired up with us.

Operator picks