Skip to main content
hc
Submit
mcpSource-backedReview first Safety ✓ Privacy ✓

Colab MCP Server

Google Colab MCP server that bridges a local MCP client to a Colab browser session through a localhost websocket proxy, letting Claude work with tools exposed by the connected Colab runtime.

by Google Colab·added 2026-06-06·
Claude CodeClaude Desktop
HarnessClaude CodeClaude Desktop
Review first — review before installing

Open the source and read safety notes before installing.

Safety notes

  • Colab MCP runs a local websocket proxy and waits for a Google Colab browser session to connect before proxying session tools.
  • The server creates a bearer-style proxy token and accepts a single authorized Colab-origin websocket connection at a time.
  • A connected Colab runtime can execute code and interact with notebooks, outputs, files, variables, packages, and mounted resources depending on the session and tools exposed by Colab.
  • Use disposable notebooks or reviewed development sessions before allowing Claude to run code or inspect outputs.
  • Avoid connecting notebooks that have access to production data, cloud credentials, private datasets, paid accelerators, or long-running jobs unless the workflow has explicit approval.

Privacy notes

  • Notebook code, prompts, outputs, logs, variables, datasets, file paths, runtime metadata, package lists, browser session details, and generated websocket tokens can be visible to the MCP client and model provider.
  • Colab notebooks may contain API keys, OAuth tokens, mounted Drive paths, secrets in environment variables, private model weights, customer data, or unpublished research.
  • Local log files are created under a temporary Colab MCP log directory by default unless a log directory is specified.
  • Review Google Colab, MCP client, model provider, and organization retention policies before sending notebook or runtime context to an assistant.

Prerequisites

  • uv or uvx available to run the server from the GitHub repository.
  • Local MCP client that supports `notifications/tools/list_changed`.
  • Google Colab account and browser session you are authorized to use.
  • Local machine access, because upstream documents that the MCP client must run locally.
  • Review of notebook, runtime, file, credential, and data access before connecting an agent to a Colab session.

Schema details

Install type
cli
Troubleshooting
No
Source repository stats
Scope
Source repo
Collection metadata
Estimated setup
15 minutes
Difficulty
advanced
Full copyable content
{
  "mcpServers": {
    "colab-mcp": {
      "command": "uvx",
      "args": ["git+https://github.com/googlecolab/colab-mcp"],
      "timeout": 30000
    }
  }
}

About this resource

Content

Colab MCP Server is a Google Colab project that bridges a local MCP client to a Google Colab browser session. It starts a local FastMCP server, opens a localhost websocket proxy, waits for an authorized Colab-origin connection, and then proxies the tools exposed by the connected Colab session to the MCP client.

Use it when Claude needs to work with a Colab notebook or runtime from a local MCP-capable client. It is useful for notebook exploration, code execution, runtime debugging, and interactive data-science workflows where the human wants Claude grounded in an active Colab session.

Source Review

These sources were reviewed on 2026-06-06. Prefer the live repository, README, license, package manifest, MCP entrypoint, session proxy, websocket server, and proxy tests for current installation and behavior details.

Features

  • Run a local FastMCP server for Colab workflows.
  • Open a browser connection to a Google Colab session.
  • Proxy a connected Colab session over a localhost websocket.
  • Use bearer token authorization and Colab-origin checks for websocket connections.
  • Accept one connected Colab session at a time.
  • Notify compatible MCP clients when the proxied tool list changes.
  • Log Colab MCP activity to a temporary log directory by default.
  • Install directly from GitHub with uvx.

Installation

Add the GitHub-backed server command to a local MCP client:

{
  "mcpServers": {
    "colab-mcp": {
      "command": "uvx",
      "args": ["git+https://github.com/googlecolab/colab-mcp"],
      "timeout": 30000
    }
  }
}

The upstream README notes that the MCP client must run locally and support notifications/tools/list_changed. After starting the MCP server, use the connection tool to open the Colab browser connection and attach the session.

Use Cases

  • Let Claude inspect or work with an active Colab notebook session.
  • Run notebook code with a human watching the connected runtime.
  • Debug Python, package, data-loading, or runtime issues in Colab.
  • Explore notebook outputs and generated artifacts from a local MCP client.
  • Coordinate interactive data-science or machine-learning experiments where Colab is the execution environment.

Safety and Privacy

Colab MCP can connect an assistant to a live notebook runtime. Treat the connected session as code execution with access to whatever the notebook, runtime, browser session, and mounted resources can reach.

Use clean notebooks, temporary runtimes, and test data for demos. Avoid connecting sessions with production datasets, private Drive mounts, cloud credentials, API keys, customer records, unpublished research, paid accelerator jobs, or long-running workloads unless the access has been reviewed.

The local proxy uses websocket authorization and origin checks, but notebook data can still flow into the MCP client, model transcript, logs, and downstream tools. Review local log files and connected client retention settings before using the server with sensitive work.

Duplicate Check

Existing entries cover Google Workspace, Google Cloud, notebook-style research tools, and Python execution-adjacent servers, but no Colab MCP entry, googlecolab/colab-mcp, or matching source URL was found in content/mcp.

#colab#notebooks#python#runtime#google

Source citations

Signals

Loading live community signals…

More like this, weekly

A short, calm digest of reviewed Claude resources. Unsubscribe any time.