Colab MCP Server
Officialby Google (googlecolab)
Google's official open-source MCP server for Google Colab, released March 2026. Connects any MCP-compatible AI agent — Claude Code, Gemini CLI, or custom frameworks — directly to Colab notebook runtimes with GPU access. Agents can create notebooks, write and execute Python code, manage dependencies, and organize cell structure without leaving their local workflow.
developmentgoogle-colabmcpjupyterpythongpunotebook-automationmodel-context-protocol
Installation
npx git+https://github.com/googlecolab/colab-mcp
Key Features
- ✓Create and manage Google Colab notebooks from any MCP-compatible AI agent
- ✓Write Python code cells and execute them in real-time within a Colab runtime
- ✓Access free GPU runtimes (T4, A100) for ML training and inference tasks
- ✓Install dependencies via !pip install commands through the agent interface
- ✓Organize and rearrange notebook cells programmatically
- ✓Bridge local AI workflows with Colab's cloud compute environment
Use Cases
- →Run ML training jobs on free Colab GPUs from Claude Code without switching context
- →Build and execute data analysis notebooks via natural language instructions
- →Train or fine-tune models using Colab's A100 runtimes from your terminal
- →Generate reproducible research notebooks with automatic code execution
- →Use Colab as a sandboxed Python environment for agent-driven experiments
FAQ
Server Stats
GitHub Stars
325
Updated
3/22/2026
NPM Package
git+https://github.com/googlecolab/colab-mcp