Colab MCP
colab-mcp
The official Google Colab MCP server: connects a local agent to a Colab session in the browser so code runs on its resources
- 1,222 ★ on the repository
- Last commit
- Apache-2.0
Broad access
Why this level
- Runs code in your Colab session
- The agent gets access to the session's files
Install
- Works in
Manual install
{
"mcpServers": {
"colab-mcp": {
"command": "uvx",
"args": ["git+https://github.com/googlecolab/colab-mcp"],
"timeout": 30000
}
}
}The block from the README. uv is needed: pip install uv. The client must be local and support notifications/tools/list_changed.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- Runs code in your Colab session
- The agent gets access to the session's files
This is third-party code. Review the repository files before installing.
What it does
The server connects a local agent to a Google Colab session open in the browser. The client must run locally and support the `notifications/tools/list_changed` notification. It launches through `uvx`, which needs `uv` (`pip install uv`), and the exact client config block is in the README. The project is written in Python, version 1.0.2, and the repository belongs to the Google Colaboratory organization.
Who it is for. For analysts and researchers who want an agent to run code on Colab resources, including accelerators, instead of on their computer.
Good fit when
- You need to run heavy code on Colab resources from an agent
- You need to work with a Colab notebook open in the browser
- You need an official bridge between a local agent and Colab
Not a fit when
- The client does not support notifications/tools/list_changed
- You are not ready to let an agent run code in your Colab session
Example request
Run training of a small model on this dataset in my Colab session and show the loss curveMCP
- Transport
- stdio
- Authentication
- not required
Limitations
The server runs code in your Colab session, so the agent gets wide access to it and its files. The client must be local and support list_changed. uv and a Colab session in the browser are needed. Availability of Google from Russia is not stated in the documentation.
How to disable. Remove the colab-mcp entry from the client config and close the Colab session.
FAQ
What does Colab MCP do?
It connects a local agent to a Google Colab session in the browser and lets it run code on its resources.
What does the client need?
To be local and support notifications/tools/list_changed.
How does it launch?
Through uvx, which needs uv: pip install uv. The config block is in the README.
Discussion
No questions or reviews yet.
Related
MCP serversDatalayer's MCP server for Jupyter: the agent opens notebooks, edits cells and runs code in a live JupyterLab session
MCP server for Google Sheets and Google Drive: an agent reads, edits and creates spreadsheets, tabs and charts
MCP server and CLI to drive real Microsoft Excel: Power Query, DAX, VBA, PivotTables and charts
BioMCP
biomcp
One binary and a single command grammar for 70 biomedical sources: PubMed, ClinVar, ClinicalTrials.gov, OncoKB. MCP mode starts with the serve command
Official MongoDB MCP server: queries and aggregations, index management, Atlas clusters and local deployments
The official MCP server for the Chroma vector database: collections, documents, semantic and full-text search
Comparison with similar tools
Entries that share the main search query, and how they differ.
| Entry | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| Colab MCP (this entry) | yes | not required | 1,222 | |
| Run Gemini CLI | yes | 2,105 | an official Google action for GitHub Actions | |
| Gemma Skills | yes | 1,009 | an official Google skill for developing on Gemma |
README in short
The README describes a bridge between a local agent and a Colab session, the client requirements and launching through uvx. Apache-2.0 license.