Microsoft's Data Formulator: The AI Analyst Tool You Probably Missed
It’s called data-formulator.
Connect any source, CSV, Postgres, Bigquery, even a live URL, then build charts with a mix of drag-and-drop and plain english.
The AI agent writes the SQL and transforms underneath and hands you a chart you can actually edit, not a code dump.
- drag fields onto x/y/color and it builds the chart
- anchor a cleaned result so follow-ups don't drift back to raw data
- branch any chart to explore a variation
- connect live data with auto-refresh
Fact-check on the claims:
True, but not brand new. Data Formulator is a real Microsoft project (Microsoft Research), open-source on GitHub under MIT license — but it wasn't "just" open-sourced. It initially released back in October 2024 and has been actively updated since, with the latest release (0.5.1) landing in December 2025. So it's an established project getting continuous updates, not a fresh drop.
Data sources — accurate. Community-contributed data loaders support Google BigQuery, MySQL, and Postgres, plus earlier loaders for MySQL, Azure Data Explorer, Azure Blob, S3, and MongoDB. CSV/file upload has been supported from the start. I didn't find explicit confirmation of a generic "live URL" connector, though the loader system is extensible enough that it's plausible.
Drag-and-drop + plain English — accurate. It combines UI interactions and natural language inputs rather than requiring users to describe everything in natural language, making it easier to design charts while delegating data transformation to AI.
AI writes SQL/transforms — accurate. Users can work with AI agents that generate SQL queries to transform data and create visualizations, running on DuckDB under the hood for larger datasets.
Anchoring — accurate. This is a real, named feature: anchoring an intermediate dataset so follow-up analysis builds on the anchored data rather than drifting back to the original, which helps keep AI agents from getting confused.
Branching — accurate. The tool supports anchoring a result, following up, creating a new branch, or joining tables as part of its workflow.
Auto-refresh — unverified. I couldn't find documentation confirming a live-refresh feature for connected data sources; worth double-checking directly on the repo/docs before repeating that claim.
One more nuance worth flagging: Microsoft itself describes this as an application from Microsoft Research that uses large language models to transform data and calls it a research prototype rather than a finished product — so "one of the best AI data analysis tools out there" is your editorial framing, not Microsoft's, and worth softening or attributing as opinion if you're posting this.