Bring Georgia's public data straight into Claude, ChatGPT, and Gemini.

What it is, in plain language

A connector, sometimes called an MCP server (Model Context Protocol), is a small bridge between your AI assistant and a source of real information. Once it is connected, when you ask about Georgia data the assistant looks up the actual cleaned datasets rather than relying on whatever was in its training.

It serves the same standardized data as the dashboard and the API on this site, but inside Claude, ChatGPT, or Gemini, as part of a normal conversation.

How it helps

• Answer questions from real data. Pull graduation rates, enrollment, test scores, and more, filtered by year, place, and demographics.
• Compare places and years. Look at districts, schools, and counties side by side, or track a single metric over time.
• Link across datasets. Join across topics on shared geography and demographic keys, with more domains (such as criminal justice) coming.
• Keep the provenance. Answers come from versioned, documented datasets with clear sources, not from guesses.

Why this beats asking an AI directly

• Straight from the cleaned data. It reads our standardized datasets, not whatever happened to be in the model's training data.
• No made-up numbers. It returns the data that exists, or tells you when a value is suppressed or unavailable, instead of inventing figures.
• Cross-dataset linking built in. Shared keys let it join datasets that would otherwise take custom data engineering.
• Always current. Newly published datasets appear automatically, with no change on your end.

Things you can ask once it's connected

• “Show graduation rates by district in metro Atlanta over the last five years.”
• “How have ACT scores changed in this district over the past decade?”
• “Which schools in Fulton County have the lowest ELA proficiency rates?”
• “Compare enrollment by grade level across two districts.”
• “What datasets do you have, and what does each one cover?”

The tools it exposes

Under the hood the connector offers a small set of tools your assistant calls for you. You do not need to learn them, but here is what it can do:

• Find datasets. List and search the full catalog of topics and dimensions.
• Inspect a dataset. Read its columns, filter keys, and the valid values for each filter.
• Query and aggregate. Pull filtered rows, or get server-side averages, sums, and counts grouped by year, place, or category.
• Resolve labels. Look up district, school, and demographic names from the shared dimension tables.
• Read the contract. Fetch a dataset's full machine-readable schema and documentation.
• Link across datasets. Discover and run guarded, read-only cross-dataset joins.

How to connect it

Connecting takes a minute. The connector URL is the same everywhere:

{
  "mcpServers": {
    "georgia-civic-data": {
      "httpUrl": "https://mcp.georgiacivicdata.org/mcp"
    }
  }
}

Good to know

• It only looks things up. The connector reads public data. It does not change anything and does not store your conversations.
• No sign-in required. The data is public, so the connector itself needs no account or key. A paid AI plan may still be required to add custom connectors.
• Not a government service. This connects to Georgia Civic Data, an independent, non-partisan project. It is not affiliated with the State of Georgia or any state agency.