home / mcp / axion planetary mcp server

Axion Planetary MCP Server

World's first Virtual Satellite that you can connect with MCP

Installation

Add the following to your MCP client configuration file.

Configuration

{
  "mcpServers": {
    "dhenenjay-axion-planetary-mcp": {
      "url": "https://axion-mcp-sse.onrender.com",
      "headers": {
        "PORT": "3000",
        "AWS_REGION": "us-east-1",
        "REQUIRE_AUTH": "true",
        "AXION_API_KEY": "your-key-here",
        "AWS_ACCESS_KEY_ID": "<AWS_ACCESS_KEY_ID>",
        "AWS_SECRET_ACCESS_KEY": "<AWS_SECRET_ACCESS_KEY>",
        "AXION_S3_EXPORTS_BUCKET": "<AXION_S3_EXPORTS_BUCKET>"
      }
    }
  }
}

Axion MCP provides a flexible, container-ready bridge between conversational agents and satellite imagery platforms. It enables hosted or self-hosted access to AWS-backed geospatial data, with tools for searching, processing, and classifying imagery, all controllable through simple natural language prompts. This makes Earth observation approachable for individuals and teams who want quick insights from satellite data without heavy infrastructure setup.

How to use

You can use Axion MCP from either a hosted service or a self-hosted setup. In the hosted mode, you connect your client (for example Claude Desktop) to the remote MCP URL and issue natural language queries like “Show me NDVI vegetation health for Iowa farmland” to retrieve maps, indices, and analysis results. In a self-hosted setup, you run the MCP server locally or in your cloud account and connect your client to the local SSE endpoint via a bridge, then perform the same kinds of queries.

How to install

Prerequisites: you need Node.js 18+ (for local/self-hosted setups) or Docker (recommended for easiest startup). Satellite data is accessed from free public APIs.

1) Hosted option: no installation required. Obtain an API key and configure your client to point at the hosted MCP URL.

2) Self-hosted Docker option (recommended for local use): open a terminal and run the following commands to clone, build, and run the MCP server.

# Clone this repository
git clone https://github.com/Dhenenjay/Axion-Planetary-MCP
cd Axion-Planetary-MCP

# Build the Docker image
docker build -t axion-mcp .

# Run without authentication (for local/personal use)
docker run -p 3000:3000 -e REQUIRE_AUTH=false axion-mcp

# OR run with authentication (for production/shared use)
docker run -p 3000:3000 \
  -e REQUIRE_AUTH=true \
  -e AXION_API_KEY=your-secret-key-here \
  axion-mcp

# Test it's working
curl http://localhost:3000/health
# Should return: {"status":"healthy","tools":6}

3) Self-hosted from source (alternative to Docker):

Clone the repository, install dependencies, build the SSE server, and start the service.

# Clone and install
git clone https://github.com/Dhenenjay/Axion-Planetary-MCP
cd Axion-Planetary-MCP
npm install

# Build the SSE server
npm run build:sse

# Run (no auth)
REQUIRE_AUTH=false npm run start:sse

# OR with auth
REQUIRE_AUTH=true AXION_API_KEY=your-key npm run start:sse

Connecting Claude Desktop to your MCP server

If you are using the hosted server, simply configure Claude Desktop to connect to the hosted URL and start asking questions. If you are running locally, use a bridge to connect Claude Desktop to your local SSE endpoint.

Available tools

axion_data

Search satellite imagery from Sentinel-2, Landsat, and NAIP across global and US-specific collections.

axion_map

Generate interactive maps with vegetation and water indices, true/false color composites, and custom band expressions.

axion_process

Process imagery to create cloud-free composites, temporal mosaics, and band calculations.

axion_classification

ML-based land cover classification using a Random Forest model with training points and accuracy metrics.

axion_export

Export results as GeoTIFFs, PNG thumbnails, map tiles, or cloud storage (S3).

axion_system

Provide system health checks, configuration info, and available data collections.