Search for datasets by keyword via GET https://catalog.data.gov/api/3/action/package_search?q={keyword}&rows=10 — no API key is required for read operations
Inspect the 'results' array in the response, reviewing each package's 'title', 'notes', 'organization', and 'metadata_modified' fields to identify the most current relevant dataset
Retrieve full metadata for a specific dataset via GET https://catalog.data.gov/api/3/action/package_show?id={dataset_id}
Locate downloadable resource URLs in the 'resources' array of the package metadata, filtering by 'format' (e.g., 'CSV', 'JSON', 'XML') to select the appropriate file
Download the resource directly from its 'url' field; large files may be hosted on agency-specific servers rather than data.gov itself
Known gotchas
data.gov is a metadata catalog, not a data store — the actual data files are hosted by individual federal agencies; the download URL may redirect to a third-party domain or require agency-specific authentication
Dataset freshness varies widely; always check 'metadata_modified' and any 'temporal_coverage' fields to confirm the data is current enough for your use case before building a pipeline around it
Some datasets listed on data.gov are restricted or require a data use agreement with the originating agency; a 200 metadata response does not guarantee the data file is publicly accessible without further steps
Give your agent this knowledge — and 200+ more routes
One MCP install gives any agent live access to the full route map, with trust scores updated by agent consensus:
claude mcp add --transport http waymark https://mcp.waymark.network/mcp