Google released an open-source <a href="https://github.com/google-colab/colab-mcp" target="_blank" rel="noopener nofollow">Colab MCP Server</a> that lets any MCP-compatible AI agent control Google Colab notebooks programmatically.
The practical upside: instead of manually copying code between a local terminal and a Colab notebook, you can point an agent like Gemini CLI, Claude Code, or a custom agent directly at Colab. From there it can create cells, write and execute Python, generate visualizations, and manage the full notebook lifecycle in the cloud.
Setup requires git, Python, and the <code>uv</code> package manager, plus a short JSON config snippet. The project is open source on GitHub and accepting contributions.

Collections

Google released an open-source MCP server for Google Colab that enables any MCP-compatible AI agent to programmatically control Colab notebooks. Agents like Gemini CLI or Claude Code can create cells, write and execute Python, generate visualizations, and manage the full notebook lifecycle without manual copy-pasting. Setup requires git, Python, and the uv package manager with a short JSON config.

Google Colab now has an MCP server for AI agent control

The link to the MCP server is not found.

this is the correct link:
<a href="https://github.com/googlecolab/colab-mcp" target="_blank" rel="noopener nofollow">https://github.com/googlecolab/colab-mcp</a>

also maybe anyone finds this interesting:
<a href="https://app.daily.dev/posts/dmBZxGgBW" target="_blank" rel="noopener nofollow">Stress Test of Colab MCP</a>