Ask your AI assistant about your community

The address

That address is all your assistant needs. It finds the sign-in on its own: you sign in with your DataWorks email and allow the connection, and it is connected.

Before you connect

1. A seat. The MCP connector is a service your organization buys from DataWorks (what we sell); its admin adds your email to its DataWorks account in the portal (each member email is a seat). No account yet? Ask us.

That is all. The seat reads the DataWorks data and Google Data Commons' data alike; no key of your own is needed.

Connect

Menu names move between releases of each app; if a step does not match what you see, the app's own help names the current place for custom connectors.

Gemini (the app)

1. In the left menu choose Customize, then Connectors, then Custom.
2. Under Add a custom app link paste the address. Leave Advanced settings (client ID, client secret) empty: the DataWorks Commons registers the connection itself.
3. Choose Next, sign in with your DataWorks email, then Allow.

Gemini then lists DataWorks Commons under its connected apps; ask as you would ask a colleague.

Gemini CLI

Add the server to ~/.gemini/settings.json (or a project's .gemini/settings.json):

{ "mcpServers": { "dataworks": { "httpUrl": "https://api.dataworkscommons.org/mcp" } } }

Then run /mcp auth dataworks in the CLI; a browser opens for the sign-in.

Claude (claude.ai and the desktop app)

1. Open Settings, then Connectors, and choose Add custom connector.
2. Name it DataWorks Commons and paste the address. Leave the OAuth client fields empty.
3. Choose Connect, sign in with your DataWorks email, then Allow.

On Team and Enterprise plans an owner may add the connector once for everyone; each member then connects with their own email.

ChatGPT

1. In Settings, under Apps & Connectors (Advanced), turn on Developer mode if your ChatGPT plan asks for it.
2. Create a connector: name DataWorks Commons, the address above, authentication OAuth.
3. Sign in with your DataWorks email and Allow. ChatGPT asks you to confirm each tool call.

Claude Code

claude mcp add --transport http dataworks https://api.dataworkscommons.org/mcp

Then run /mcp in Claude Code, choose dataworks and sign in.

VS Code and Cursor

VS Code, .vscode/mcp.json in your project (or your user settings):

{ "servers": { "dataworks": { "type": "http", "url": "https://api.dataworkscommons.org/mcp" } } }

Cursor, ~/.cursor/mcp.json:

{ "mcpServers": { "dataworks": { "url": "https://api.dataworkscommons.org/mcp" } } }

Start the server from the editor; a browser opens for the sign-in.

The MCP Inspector (for testing)

npx @modelcontextprotocol/inspector

Add a server: transport streamable-http, the address above; connect, sign in, then open Tools.

What to ask

• "What is the poverty rate and the violent crime count in Tulsa?"
• "How has median household income changed in Oklahoma County since 2015?"
• "Compare home values in Tulsa and Oklahoma City."

Your assistant uses four tools: search_variables (find measures), find_places (place names to ids), get_data (the numbers, each with its source) and describe_variable (what a measure means and where it is published).

What it can read

• DataWorks data and Google Data Commons data, on the same seat: every public measure, for any place, without limit. Each answer names its source.
• Your own place's data. An organization's Insights, its row-level data and the data built for it are read for its own place only. A measure whose source allows it on this site only is listed as not served, with the page to read it on.

If it does not connect

Questions: info@thedataworks.org. Building your own code instead? See the API guide.