Data Oakland

What Data Oakland does

The Data Oakland MCP server gives MCP clients access to the Oakland Open Data catalog at data.oaklandca.gov. It supports three dataset-focused operations:

• datasets searches the catalog by keyword and returns each matching dataset’s resource ID, name, description, category, and update date.
• metadata retrieves a dataset’s schema and related details, including columns, types, row count, category, and last-updated information.
• query runs a Socrata SoQL query against a dataset identified by its resource ID and returns matching rows as JSON.

A typical workflow starts with a catalog search, continues with metadata inspection, and then uses the returned resource ID for a targeted query. The Data Oakland MCP server is suited to agents that need Oakland-specific public datasets in a structured format.

How it works

The hosted connection uses the Pipeworx gateway. The pack endpoint is:

https://gateway.pipeworx.io/data-oakland/mcp

The endpoint directly lists the Oakland tools and also exposes shared Pipeworx gateway tools. Those shared tools can route broader questions through the catalog. The full gateway is available at https://gateway.pipeworx.io/mcp when an agent needs every pack listed directly; both gateway URLs can reach the same underlying data sources.

A local package is also available through npm. In that mode, the Data Oakland MCP server communicates over standard input and output and exposes only its own pack tools, without the gateway’s shared meta-tools or cross-catalog routing.

The query operation accepts SoQL clauses through where, select, group, and order, along with limit and offset. Clause values should be supplied without the leading $ used in some Socrata API examples. A resource ID such as ym6k-rx7a identifies the dataset being queried.

Setup and configuration

For an MCP client that supports remote endpoints, add the hosted URL as the server connection:

{
  "mcpServers": {
    "data-oakland": {
      "url": "https://gateway.pipeworx.io/data-oakland/mcp"
    }
  }
}

The package can instead run locally with npx:

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

Run npx -y @pipeworx/mcp-data-oakland directly to verify that the local process starts. HTTP clients can call the gateway’s tool API directly; for example, dataset search uses POST https://gateway.pipeworx.io/v1/tools/data_oakland_datasets with a JSON body such as {"query":"crime"}.

Tools and capabilities

The Data Oakland MCP server provides a compact path from discovery to analysis:

1. Search for relevant Oakland datasets by keyword.
2. Examine fields and data types before writing a query.
3. Apply SoQL selection, filtering, grouping, ordering, pagination, and row limits.
4. Receive query results as JSON for agent reasoning or downstream processing.

The gateway endpoint additionally includes shared tools such as ask_pipeworx, discover_tools, search_within, and memory-related tools. Those are gateway capabilities rather than Oakland dataset operations, and the local stdio mode does not include them.

Limitations and notes

This package is specific to Oakland Open Data and depends on dataset resource IDs and the Socrata query model. It does not describe a general-purpose database connection. The hosted pack endpoint currently includes shared gateway tool schemas in addition to the three Oakland tools, so its reported tool count can be higher than the pack’s direct capabilities. The local package avoids those extra schemas but also does not provide gateway-wide routing.

The project is licensed under MIT.