Starwell: World Data & Statistics

What Starwell: World Data & Statistics MCP server does

Starwell: World Data & Statistics MCP server gives AI clients access to official statistical data from 28 agencies. Listed sources include FRED, Statistics Canada, the U.S. Bureau of Labor Statistics, the World Bank, SEC EDGAR, the U.S. Treasury, the European Central Bank, the UK Office for National Statistics, INSEE, the Australian Bureau of Statistics, the OECD, Eurostat, and national statistical offices across Europe and elsewhere.

The service is intended for questions that need identifiable source data rather than uncited generated values. Observations include provenance, and series include a verification status supported by checks against live-source reference values. Open data reads do not require an API key. Computed answers, deeper analysis, and monitor operations require a free Starwell key.

How it works

The npm package is a thin stdio adapter for Starwell's remote MCP service. Its tool set follows the remote server, so changes to available sources and tools do not require a package release. Clients that support remote MCP can connect directly to the Streamable HTTP endpoint at https://starwell.dev/api/starwell/mcp; other setups can launch the package with npx -y starwell-mcp.

Starwell: World Data & Statistics MCP server returns data from openly licensed official sources. Responses include source attribution and licensing information. The README specifically notes public-domain U.S. federal statistical works, the Statistics Canada Open Licence, World Bank CC BY-4.0, and open licences associated with several European and Australian agencies. FRED data is accessed through its API, without endorsement or certification from the Federal Reserve Bank of St. Louis.

Tools and capabilities

The server exposes 12 tools:

• search_catalog searches the available data catalog.
• list_sources lists source agencies.
• list_datasets lists datasets.
• get_series retrieves a series.
• get_series_stats obtains statistics for a series.
• get_observations retrieves observations.
• factory_status reports factory status.
• answer produces computed, cited answers.
• deep_analysis runs deeper analysis.
• create_monitor, list_monitors, and delete_monitor manage monitors.

Together, these tools support finding a dataset, selecting a series, retrieving observations, and generating cited computed results. The material does not specify the exact input schemas or monitor scheduling behavior, so those details should be checked in the service documentation.

Setup and configuration

Remote-capable clients can use the Streamable HTTP URL directly. The README provides configurations for Claude Code, Claude Desktop, Cursor, OpenAI Codex CLI, Gemini CLI, and ChatGPT developer mode. For stdio configurations, launch npx -y starwell-mcp and register it as an MCP server.

Set STARWELL_API_KEY to a bearer key from the Starwell account page when using answer, deep_analysis, or monitor tools. Leave it unset for open data reads. STARWELL_MCP_URL can replace the default endpoint, including for staging or self-hosted deployments.

Limitations and notes

The package depends on the remote Starwell service, and its tool surface can change as that service changes. Only openly licensed official sources are served. The provided material does not document rate limits, historical coverage for individual datasets, response schemas, or the behavior of each agency-specific connector. Verify those details before building workflows that depend on a particular series or source.