jupyter-earth

🪐 🌎 Jupyter Earth MCP Server

Installation

Run this command in your terminal to add the MCP server to Claude Code.

Run in terminal:

Command

claude mcp add --transport stdio datalayer-jupyter-earth-mcp-server docker run -i --rm -e SERVER_URL -e TOKEN -e NOTEBOOK_PATH --network=host datalayer/jupyter-earth-mcp-server:latest \
  --env TOKEN="MY_TOKEN" \
  --env SERVER_URL="http://localhost:8888" \
  --env NOTEBOOK_PATH="notebook.ipynb"

How to use

Jupyter Earth MCP Server exposes a focused set of MCP capabilities to enable geospatial analysis workflows directly from Jupyter notebooks. Currently it offers a single built-in tool, download_earth_data_granules, which lets you fetch Earth science data granules from NASA Earthdata and save them to a local folder via a notebook cell. There is also a prompt workflow named download_analyze_global_sea_level that guides you through downloading and preparing global sea level data for analysis. To use the server, run the MCP server container with the provided environment variables (SERVER_URL, TOKEN, and NOTEBOOK_PATH) and point your MCP client or Claude Desktop integration at the configured SERVER_URL and TOKEN. Inside a notebook, call the download_earth_data_granules tool to retrieve data, then continue analysis with your existing Jupyter workflow. The prompts help you initiate higher-level data retrieval and analysis tasks without constructing raw MCP requests manually.

How to install

Prerequisites:

Docker installed on your machine (community edition is fine).
Access to the internet to pull the MCP server image.
A Jupyter environment where you will run notebooks that interact with the MCP server.

Installation steps:

Install Docker (follow platform-specific instructions at https://docs.docker.com/get-dstarted/).
Pull the MCP server image (optional if you rely on Claude config to run the server container): docker pull datalayer/jupyter-earth-mcp-server:latest
Prepare your environment variables for the MCP channel:
- SERVER_URL: the endpoint where your MCP server is reachable (e.g., http://localhost:8888).
- TOKEN: a secret token used for authentication with the MCP server.
- NOTEBOOK_PATH: path to the notebook you are running, relative to the container's perspective (e.g., notebook.ipynb).
Run the MCP server via Docker using the provided configuration (example for Linux/macOS/CLI): docker run -i --rm -e SERVER_URL=http://localhost:8888 -e TOKEN=MY_TOKEN -e NOTEBOOK_PATH=notebook.ipynb --network=host datalayer/jupyter-earth-mcp-server:latest
In your MCP client (or Claude Desktop), configure the MCP server entry named jupyter-earth with the docker-based command and the same environment variables.

Optional: If you want to use the built-in Makefile targets, you can build the image locally if provided by the repo: make build-docker

Remember to adjust SERVER_URL to match how you expose the MCP server in your environment (host vs container networking).

Additional notes

Tips and common issues:

Ensure the SERVER_URL in both the MCP client and the container matches the address you expose from your host or CI environment.
When running in Docker, if you need access to local notebooks, consider using --network=host (Linux) or appropriate port mappings for your OS. If you cannot use host networking, map the container ports to your host and update SERVER_URL accordingly.
The tool download_earth_data_granules expects you to provide notebook-ready inputs within your cell, including folder_name, short_name, count, and optional temporal and bounding_box parameters.
If you encounter authentication issues, verify that TOKEN is correctly passed to the container and that the MCP client is sending the token with requests.
This MCP server is labeled as archived in the README; for longer-term use, consider migrating to earthdata-mcp-server if you need updated features or security patches.
The notebook path (NOTEBOOK_PATH) should be relative to where JupyterLab is started or the container’s working directory, depending on your deployment configuration.