Vibeleaderboard
← Back to Tools
Index / tool

MCP Jupyter

https://github.com/block/mcp-jupyter
Visit github.com
Category
AI Agents
Rank
No. 995Tools index
Pricing
Open Source
Type
TOOL
Builder
block
GitHub
45 stars
Latest release
v2.0.2
Date

About

MCP server that gives AI agents the ability to control and execute code in Jupyter notebooks.

What it does

MCP Jupyter bridges compatible assistants with a running JupyterLab environment. It edits notebook cells through Jupyter’s REST interface, executes code against an existing kernel, and preserves variables between human and agent turns. Optional real-time collaboration keeps browser edits synchronized automatically.

Why it's ranked here

A strong choice for collaborative notebook work because it combines explicit REST-based saves, persistent kernel state, and focused notebook operations. Its value is narrower than general automation: it assumes an existing Jupyter setup, and its evolving interface may break between releases.

What's good

Four consolidated tools cover reading, setup, cell changes, execution, and package installation without exposing a sprawling command set. State hashing detects notebook changes before dependent edits. It supports local standard input communication plus stateful or stateless HTTP, and includes unit, integration, and language-model tool-call testing.

Tradeoffs

You must install UV, run an accessible JupyterLab server, and configure an MCP-compatible client. Automatic browser synchronization needs the collaboration extension; without it, users must save their edits and reload agent changes manually. The project explicitly offers no compatibility guarantees between versions, so upgrades may require client adjustments.

How to use it well

Use it for iterative data exploration, modeling, debugging, and reports where a person and assistant alternate inside one live notebook. Give specific, incremental requests and distinguish code cells from markdown cells. It does not replace JupyterLab setup, provide a hosted notebook environment, or serve as a general-purpose automation layer beyond notebooks.

Technical notes+

pyproject.toml defines a Python 3.10+ Hatchling package and the mcp-jupyter console entry point. src/mcp_jupyter/__init__.py selects stdio or streamable HTTP and supports host, port, and stateless HTTP options. src/mcp_jupyter/server.py builds the FastMCP server, routes four consolidated tools, and connects execution to existing kernels. src/mcp_jupyter/rest_client.py performs notebook reads and explicit saves through Jupyter REST endpoints. src/mcp_jupyter/state.py stores a class-level SHA-256 content hash, waits 1.5 seconds before fetching content, and rejects state-dependent work after detected changes. src/mcp_jupyter/utils.py removes base64 image payloads from returned outputs while retaining text indicators.

Observed

Primary language
Python
Runtime requirement
Python 3.10 or newer
Packaging
Hatchling build backend with a console entry point
Installation surface
Runs as uvx mcp-jupyter; UV is required
Interfaces
MCP server using stdio or streamable HTTP transport
HTTP modes
Supports session-based and stateless HTTP operation
Testing structure
Documented unit, integration, and LLM tool-call tests under tests/

Read from README.md, pyproject.toml, src/mcp_jupyter/state.py, src/mcp_jupyter/utils.py, src/mcp_jupyter/server.py, src/mcp_jupyter/__init__.py, src/mcp_jupyter/__main__.py, src/mcp_jupyter/notebook.py, src/mcp_jupyter/rest_client.py, docs/README.md, docs/docs/intro.md, docs/docs/usage.md, docs/docs/quickstart.md, docs/docs/development.md, docs/docs/architecture.md.

What it can do

  • Execute Python code in Jupyter notebooks

    Python code snippets or scripts → Code execution results and outputs

  • Create new Jupyter notebook cells

    Code or markdown content → New notebook cells with specified content

  • Read and analyze existing notebook content

    Jupyter notebook files → Notebook structure, cells, and content data

  • Modify existing notebook cells

    Cell identifiers and new content → Updated notebook cells with modified content

  • Run data analysis workflows

    Datasets and analysis instructions → Analysis results, visualizations, and insights

  • Generate and execute machine learning code

    ML requirements and data specifications → Trained models, predictions, and evaluation metrics

  • Create data visualizations and plots

    Data and visualization specifications → Charts, graphs, and visual representations

Tags

mcpjupyternotebooksai-agentpython

Tech Stack

Python

Comments (0)

No comments yet

Editorially curated, with community endorsements as a secondary signal. Corrections welcome.