ERDDAP

You can access ERDDAP oceanographic and environmental data through two MCP implementations that connect to ERDDAP servers: a local stdio server for use on your machine and a remote HTTP server for cloud deployments. Use Claude Desktop to run these MCP servers and perform common data discovery and retrieval tasks with ease.

How to use

You use an MCP client inside Claude Desktop to interact with ERDDAP data. Start the local stdio MCP server on your machine for a straightforward setup, or use the remote HTTP MCP server when you need cloud access and shared usage. The local server runs without network dependencies, while the remote server is accessed via the mcp-remote proxy to enable HTTP communication. Once configured, you will have four core tools to discover datasets, inspect metadata, and download data for analysis.

Begin by choosing the appropriate server configuration in Claude Desktop so the tools load automatically on startup. The local server uses Python to run a script that exposes four data access tools. The remote server uses an HTTP-based MCP endpoint that works behind the mcp-remote proxy for secure remote access.

How to install

Prerequisites: you need Python and Node.js (for the remote proxy) installed on your system.

{
  "mcpServers": {
    "erddap-local": {
      "command": "python",
      "args": ["/path/to/erddapy_mcp_server.py"]
    },
    "erddap-remote": {
      "command": "npx",
      "args": ["mcp-remote", "https://erddap2mcp.fly.dev/"]
    }
  }
}

Additional notes and setup details

VPN Usage: Disable any VPN before using ERDDAP MCP tools. VPNs can cause 404 errors, incomplete downloads, or timeouts.

Local MCP server setup steps are designed to be simple and self-contained, with no network dependencies. Remote access requires the mcp-remote proxy to route HTTP requests from Claude Desktop to the remote MCP server.

You can run both local and remote servers simultaneously to compare results or to provide a dependable workflow that supports multiple users.

Configuration and usage tips

Commonly used actions include listing well-known ERDDAP servers, searching datasets by keyword, retrieving dataset metadata, and downloading data for inspection in a pandas DataFrame.

If you plan to share results or collaborate, prefer the remote HTTP server with the mcp-remote proxy, while keeping the local server as your isolated testing ground.

Testing your setup

After configuring Claude Desktop, restart the application and verify that the ERDDAP tools appear in the tool list. If a tool does not load, check that the corresponding MCP server is running and that the CLI/GUI configuration points to the correct command and URL.

Available tools

list_servers

Show well-known ERDDAP servers available for discovery and selection.

search_datasets

Search for datasets by keyword and retrieve matching results.

get_dataset_info

Fetch detailed metadata about a specific dataset.

to_pandas

Download and preview data in a pandas-friendly format for quick analysis.