Datos Co

What Datos Co MCP server does

Datos Co MCP server gives MCP-compatible agents access to datasets published on Colombia’s national open-data portal, datos.gov.co. The portal uses the Socrata platform, and datasets are identified by four-by-four codes such as gt2j-8ykr.

The server supports a discovery-to-query workflow. An agent can search the catalog, inspect the available fields for a selected dataset, and then retrieve filtered, grouped, sorted, or aggregated rows. Dataset names and descriptions returned by discovery are in Spanish, while search terms may be written in Spanish or English.

How it works

Start with search_datasets and provide a topic such as covid, presupuesto, or contratos. The response includes dataset names, descriptions, and Socrata IDs. Pass the selected ID to dataset_columns to learn the field names, labels, and data types available for querying.

Use query_dataset after inspecting the columns. Its SoQL parameters include $select for fields or aggregates such as count(*), $where for SQL-like filtering, $order for sorting, $group for grouped results, and $q for full-text matching across rows. Returned records are arrays of objects keyed by field name. The default result limit is 50 rows, while Socrata permits up to 50,000 rows per page.

Setup and configuration

The hosted MCP endpoint is:

https://gateway.pipeworx.io/datos-co/mcp

Add that URL to an MCP client that supports remote servers. The repository names Claude Desktop, Cursor, and Windsurf as examples. The same gateway also exposes a general Pipeworx endpoint at https://gateway.pipeworx.io/mcp, which lists tools from all available data-source packs directly.

For local use, run the package with npx:

{
  "mcpServers": {
    "datos-co": {
      "command": "npx",
      "args": ["-y", "@pipeworx/mcp-datos-co"]
    }
  }
}

The local process communicates over stdin and stdout and exposes only the Datos Co tools. The hosted endpoint adds shared Pipeworx meta-tools, including routing and discovery functions.

Tools and capabilities

• search_datasets: Finds datasets on datos.gov.co using Spanish or English terms.
• dataset_columns: Reports field names, human-readable labels, and data types for a dataset.
• query_dataset: Runs SoQL queries and returns row objects.
• HTTP tool access: Hosted tools can also be called with HTTP, including the datos_co_search_datasets tool.
• ask_pipeworx: On the gateway connection, natural-language questions can be routed to the appropriate tool.

Limitations and notes

The dataset ID must come from the Socrata catalog and is required by both dataset_columns and query_dataset. Field inspection should come before query construction so that $select and $where use valid names. The hosted pack endpoint includes shared gateway tools in addition to the three pack-specific tools, so its complete tool list is larger than the local server’s list. The local package does not provide those shared meta-tools or gateway routing.