Easy Jupyter Editor MCP
A lightweight Model Context Protocol (MCP) server designed to allow AI Agents to safely and structurally edit Jupyter Notebook (.ipynb) files.
Background
This MCP server was developed to enable AI agents to safely manipulate Jupyter Notebook (.ipynb) files.
In many AI agent environments (such as Antigravity), directly executing notebook files or editing complex .ipynb JSON structures as raw text can be difficult due to security restrictions or the risk of corrupting the file format.
This project aims to "structurally edit Notebook files correctly while bypassing execution permission issues." By using the nbformat library, it allows adding, editing, and deleting cells without breaking the JSON structure.
Features
Provides the following tools:
read_notebook: Retrieves a list of cells (index and content summary) from a Notebook.get_cell: Retrieves the full source code of a specific cell by index.edit_cell: Modifies the content of a specific cell.add_cell: Adds a new code or Markdown cell.delete_cell: Deletes a specific cell.create_notebook: Creates a new, empty Notebook.
Installation & Usage
It is recommended to run this server using uv.
1. Claude Desktop / MCP Client Configuration
Add the following configuration to claude_desktop_config.json.
Using PyPI (Recommended)
This is the simplest way to run the server.
{
"mcpServers": {
"easy-jupyter-editor": {
"command": "uvx",
"args": ["easy-jupyter-editor-mcp"]
}
}
}Running directly from GitHub
{
"mcpServers": {
"easy-jupyter-editor": {
"command": "uv",
"args": [
"run",
"--from",
"git+https://github.com/YourUsername/easy-jupyter-editor-mcp",
"easy-jupyter-editor-mcp"
]
}
}
}Running Locally for Development
{
"mcpServers": {
"easy-jupyter-editor": {
"command": "uv",
"args": [
"run",
"--with",
"mcp[cli]",
"--with",
"nbformat",
"python",
"/absolute/path/to/easy-jupyter-editor-mcp/src/easy_jupyter_editor_mcp/__init__.py"
]
}
}
}Development
# Build
uv build
# Publish to PyPI
uv publish
