NYC School of Data 2026

"Large language models like ChatGPT and Claude can now connect directly to NYC Open Data through a tool called Model Context Protocol (MCP)—meaning you can ask questions in plain English and get answers informed by data pulled live from city datasets, not AI-generated guesses.

This session is a practical walkthrough of how that works, and why it matters. We'll start with a live demo of Civic AI Tools (civicaitools.org), which lets you compare LLM responses with and without live data access side by side. Then we'll go under the hood, showing how MCP connects AI tools directly to NYC Open Data's Socrata API—and how you can set up and use these tools yourself in GitHub Codespaces.

Using NYC 311 service request data as our running example, we'll show how to go from a plain-English question (""What are the top complaint types in my community district?"") to a trustworthy, citeable answer—step by step. We'll show how live data access dramatically improves the quality of AI-generated answers—and why critical thinking is still essential when evaluating any LLM output.

By the end of the session, you'll be ready to set up and use these tools on your own NYC Open Data questions.
This session pairs with Part 2 (12:15–1:15pm), where three working journalists will show you how to take the data you've queried and turn it into polished charts and visualizations using the Datawrapper MCP. Each session stands on its own, but together they cover a complete workflow from question to published chart."