rasdaman

The Rasdaman MCP server exposes Rasdaman functionality to a large language model via the MCP protocol. It translates natural language prompts into Rasdaman WCS/WCPS queries and returns results to the caller. With this server, you can list available datacubes, retrieve metadata for a specific datacube, and execute WCPS-like queries to perform analytics or produce new imagery. Use the provided tools through either a standard input/output (stdio) integration or via an HTTP API, depending on how you launch the server.

To use, configure the Rasdaman connection either through environment variables (RASDAMAN_URL, RASDAMAN_USERNAME, RASDAMAN_PASSWORD) or via the command-line options when starting rasdaman-mcp. Once started, the MCP server exposes tools such as list_coverages, describe_coverage, and execute_wcps_query. An AI agent can then call these tools by name, passing the required arguments, and receive structured results or data file paths suitable for downstream visualization or processing.

Prerequisites:

• Python 3.8+ and pip
• Access to install Python packages (internet connection)

Installation steps:

1. Create and activate a virtual environment (optional but recommended):

   python -m venv venv source venv/bin/activate # On Windows use: venv\Scripts\activate

2. Install the Rasdaman MCP package from PyPI:

   pip install rasdaman-mcp

3. Run the MCP server (example using stdio mode):

   rasdaman-mcp --rasdaman-url "http://localhost:8080/rasdaman/ows" --username rasguest --password rasguest

4. Alternatively, run in HTTP mode (standalone server):

   rasdaman-mcp --transport http --host 0.0.0.0 --port 8000 --rasdaman-url "http://localhost:8080/rasdaman/ows" --username rasguest --password rasguest

Note: You can also configure the connection via environment variables RASDAMAN_URL, RASDAMAN_USERNAME, and RASDAMAN_PASSWORD.

Tips and considerations:

• If you encounter authentication errors, verify that the Rasdaman URL is correct and that the provided credentials have access to the required datacubes.
• When running in HTTP mode, ensure the selected host/port are exposed to your environment and any firewalls allow inbound connections.
• The MCP tools exposed by rasdaman_mcp include: list_coverages(), describe_coverage(coverage_id), and execute_wcps_query(wcps_query). Use these to build natural language prompts that map to RS-based queries.
• For development, running in editable or source mode is fine, but for production consider containerizing the server and using a reverse proxy for the HTTP endpoint.