Colab Drive MCP
Sync Colab notebook files with a bounded local folder and Google Drive so agents can edit notebooks without running Colab cells.
Overview
colab-drive-mcp is a MCP server for the Build phase that synchronizes Colab notebook files through Google Drive and a local root without executing notebook code.
What is this MCP server?
- Read and write notebook files under a root directory you define (COLAB_MCP_ROOT)
- Google Drive modes: minimal file access or full discovery of existing notebooks
- Optional local OAuth client and token paths for Google Desktop flow
- Does not execute notebook code—file sync only for safer agent edits
- Stdio MCP distributed via PyPI (uvx) at version 0.1.3
- Registry version 0.1.3
- Transport: stdio
- PyPI identifier colab-drive-mcp (runtimeHint uvx)
What problem does it solve?
Agents cannot safely round-trip Colab notebooks between Drive, local git, and the IDE without manual downloads and copy-paste.
Who is it for?
Solo builders running ML or research notebooks in Colab who want Cursor or Claude Code to edit .ipynb files with Drive-backed sync.
Skip if: Users who need remote kernel execution, scheduled training jobs, or notebook runs from the agent—this server is file-only.
What do I get? / Deliverables
Notebook files stay in sync inside your declared root so agents can list, read, and update them like normal project files.
- Notebook files readable and writable within COLAB_MCP_ROOT
- Drive-linked sync per file or full discovery mode
- Agent-safe file operations without notebook execution
Recommended MCP Servers
Journey fit
Notebook file sync is part of Build when you wire ML experiments into your repo and agent workflow. It connects Google Drive, local disk, and Colab paths—classic third-party integration work, not UI or pure backend API code.
How it compares
Drive-backed notebook file MCP, not a Colab runtime or GPU orchestration skill.
Common Questions / FAQ
Who is colab-drive-mcp for?
Builders and researchers who use Google Colab notebooks but want agent-driven file edits synced via Drive and a local COLAB_MCP_ROOT.
When should I use colab-drive-mcp?
Use it during Build when integrating notebook assets into your workflow—editing, organizing, or backing up .ipynb files without executing cells.
How do I add colab-drive-mcp to my agent?
Install the PyPI package (colab-drive-mcp, uvx), set COLAB_MCP_ROOT to an absolute allowed directory, configure COLAB_MCP_DRIVE_ACCESS and Google OAuth env vars if needed, then add the stdio server to your MCP client.