Model Context Protocol high risk

Jupyter Notebook

JupyterMCP allows Claude AI to interact with and control Jupyter Notebook 6.x for AI-assisted code execution, data analysis, and visualization.

Connections & Capabilities

Connects To

GitHub

Capabilities

readwriteexecadmin

Quickstart

Install

brew install uv

Config

{
    "mcpServers": {
        "jupyter": {
            "command": "uv",
            "args": [
                "--directory",
                "/ABSOLUTE/PATH/TO/PARENT/REPO/FOLDER/src",
                "run",
                "jupyter_mcp_server.py"
            ]
        }
    }
}

Exposed MCP Tools (11)

safe

ping

Checks the connectivity to the Jupyter Notebook server.

Read-only operation with no side effects.

high

insert_and_execute_cell

Inserts a new cell at a specified position and executes its content.

Allows arbitrary code execution, potentially leading to data modification or deletion.

moderate

save_notebook

Saves the current Jupyter Notebook.

Modifies the notebook file, potentially overwriting existing content.

safe

get_cells_info

Retrieves information about all cells in the notebook.

Read-only operation that retrieves cell metadata.

safe

get_notebook_info

Retrieves information about the current notebook.

Read-only operation that retrieves notebook metadata.

high

run_cell

Runs a specific cell by its index.

Executes arbitrary code within a specific cell.

high

run_all_cells

Runs all cells in the notebook sequentially.

Executes arbitrary code in all cells, potentially causing widespread changes.

safe

get_cell_text_output

Retrieves the text output content of a specific cell.

Read-only operation that retrieves cell output.

safe

get_image_output

Retrieves the images output of a specific cell.

Read-only operation that retrieves image output.

moderate

edit_cell_content

Edits the content of an existing cell.

Modifies the content of a cell, potentially altering code or markdown.

moderate

set_slideshow_type

Sets the slide show type for a cell.

Modifies cell metadata related to slideshow presentation.

Safety Assessment

JupyterMCP offers powerful integration but poses significant risks due to its ability to execute arbitrary code without sandboxing. It's safe for controlled environments with trusted code, but risky when used with untrusted sources or without careful oversight.

Requires explicit user setup and initialization within a Jupyter Notebook.
Includes a disclaimer emphasizing the experimental nature and potential risks.
Provides example prompts to guide users on how to interact with Claude.

Allows arbitrary Python code execution, potentially leading to data modification or deletion.
Lacks sandboxing, increasing the risk of malicious code execution.
Relies on user-configured paths, which could introduce vulnerabilities if misconfigured.
Limited text output from cells could truncate important information, obscuring potential risks.