Cursor

Cursor connects to MCP servers through an mcp.json file. Use the hosted server with a header token for a zero-install setup, or the local stdio server for full read/write.

Hosted server (header token)

Create ~/.cursor/mcp.json (global) or .cursor/mcp.json in your project, and add a remote server with the Datamesh endpoint and an X-DATAMESH-TOKEN header:

{
  "mcpServers": {
    "oceanum-datamesh": {
      "url": "https://mcp.oceanum.io/datamesh",
      "headers": {
        "X-DATAMESH-TOKEN": "your-datamesh-token"
      }
    }
  }
}

Cursor treats an entry with a url (rather than a command) as a remote server. Replace your-datamesh-token with your Datamesh token. To keep the token out of the file, Cursor also supports the ${env:VAR} syntax in header values.

The hosted server is read-only and returns query_data results inline (up to ~50 MB), so large results can’t come back in the conversation. To get large data, call export_query: on the hosted server it returns a time-limited download link to the full result, which you fetch out-of-band (e.g. with curl) — the link needs no token, so treat it like a password. The local server below instead writes export_query output to a file on your machine.

Local server (stdio)

For full read/write, and to have export_query write results straight to a file on your machine (NetCDF, Parquet or CSV) rather than return a download link, run the server locally:

{
  "mcpServers": {
    "oceanum-datamesh": {
      "command": "uvx",
      "args": ["oceanum-mcp", "datamesh"],
      "env": { "DATAMESH_TOKEN": "your-token-here" }
    }
  }
}

This requires uv on your PATH. After saving, open Cursor’s Settings → MCP to confirm the server is connected and its tools are listed.

Keep your token safe

Your Datamesh token grants access to all of your organisation’s permissioned datasources — including the ability to modify or delete them. Treat it as a secret, and prefer the hosted read-only server for exploratory use.