Data Commons Agent Toolkit MCP
Official Data Commons MCP server: an agent finds statistical indicators and pulls observations for countries, regions and cities.
- 149 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- Agent queries and the API key go to the external Data Commons service.
Install
- Works in
Manual install
uvx datacommons-mcp serve stdioStdio launch for local clients. Set `DC_API_KEY` first.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Agent queries and the API key go to the external Data Commons service.
This is third-party code. Review the repository files before installing.
What it does
The `datacommons-mcp` package gives an agent open statistics from the Data Commons knowledge graph. The agent searches indicators by description, inspects variable metadata and fetches observations for one place or for all its child places, such as unemployment by county. The code defines six tools: `search_indicators`, `search_child_indicators`, `get_variable_metadata`, `get_observations`, `get_child_observations` and `get_multi_entity_observations`. Three bundled skills set the workflow: splitting a question into concepts, resolving places, choosing a variable. The server runs over stdio or streamable HTTP and supports a custom Data Commons instance.
Who it is for. Analysts and agent developers who need country and region statistics from an open source with the indicator named.
Good fit when
- An agent compares unemployment or GDP across countries.
- A value is needed for every county inside one region.
- Before a report, variable metadata is checked: what it measures and where it comes from.
- A team builds an agent on Gemini CLI or Google ADK and attaches open data over MCP.
Not a fit when
- You need private company tables that are absent from Data Commons.
- You need an indicator for the last month: coverage depends on the source.
- A Data Commons API key is not available to you.
Example request
Compare the unemployment rate across US states over recent years and name the indicator and data source.MCP
- Transport
- stdio, http
- Authentication
- API key
Environment variables
DC_API_KEYrequired, secret- Data Commons API key from apikeys.datacommons.org.
Limitations
A Data Commons API key is required; create it at `apikeys.datacommons.org` and set it as `DC_API_KEY`. Running through `uvx` needs `uv` installed. Data comes from the Data Commons graph: coverage varies by country and place level, with countries the most complete and some indicators missing at sub-national levels. Agent queries go to an external service. Documentation hints are enabled with the `X-DC-Enable-Documentation` header and work only for HTTP clients, not over stdio.
How to disable. Remove the server entry from the client settings and stop the `datacommons-mcp` process. Clear `DC_API_KEY` from the environment and revoke the key if you wish.
FAQ
Which key is needed?
A Data Commons API key from `apikeys.datacommons.org`, passed through `DC_API_KEY`.
Which transports exist?
Streamable HTTP via `uvx datacommons-mcp serve http`, port 8080 by default at `http://localhost:<port>/mcp`, and stdio for local clients.
How do I get child-place data?
For questions like unemployment in all counties of California, the server points to the `data-commons-child-places-researcher` skill and the `search_child_indicators` and `get_child_observations` tools.
Can I use my own Data Commons?
Yes, the README states support for custom instances.
Is documentation guidance on?
Not by default. An HTTP client enables it with `X-DC-Enable-Documentation: true`; it is not sent over stdio.
Discussion
No questions or reviews yet.
Related
MCP serversA DataHub MCP server: an agent searches tables and columns by metadata, traces lineage and finds analysts' SQL queries
DaData MCP
dadata-mcp
DaData MCP server with 31 tools: address suggestions, company lookup by INN, phone and passport cleaning, reference lists
Official MCP server from the OpenSearch project exposes a cluster to agents: indexes, mappings, DSL search and shard state.
An MCP server that lets an agent run local Stata: regressions, econometrics, replicating computations from papers
An MCP server with a toolbox for global briefs: markets, SEC filings, conflict, climate and news from open sources
data.gouv.fr MCP
data.gouv.fr MCP Server
Official MCP server for France's open data portal: search, explore and query datasets straight from a chat with the agent
README in short
The root README points to the server package and sample agents, with user docs at `docs.datacommons.org/mcp`. The package README covers the key, two run modes, optional documentation guidance and client setup, including Gemini CLI.