Data Formulator 0.7: AI agents meet enterprise analytics in a shared workspace

Enterprise analytics teams face a persistent problem: their data lives across disconnected systems—databases, warehouses, BI platforms, and cloud storage—while the people analyzing it often lack deep coding skills. Microsoft Research has released Data Formulator 0.7 as an open-source answer.

Three core pieces#

The system combines three capabilities designed to work together:

Data Connectors handle the plumbing. Rather than rebuild connections each time, teams set up authenticated, reusable links to databases, warehouses, BI systems, object stores, and local files. Metadata and previews live in a unified workspace, cutting integration burden for platform teams.

Context-aware agents do the analytical heavy lifting. Unlike a single-turn chat prompt, these agents see the full workspace—connected data, loaded tables, prior charts, and the user's stated goal. They can inspect data, run code in isolation, generate visualizations, and ask clarifying questions when a request is ambiguous. Results are verifiable and reproducible.

An interactive workspace lets teams iterate together. The "Data Thread" captures every question, finding, and chart in a navigable record. Users can branch into alternative analyses, compare them side by side, and revisit earlier steps. When moving from exploration to presentation, they can edit visualizations directly on a canvas or describe changes in plain language and let the agent adjust labels, colors, and layout.

The real constraint#

The release addresses a genuine friction point: most analysts aren't SQL experts, and isolated chat interactions rarely retain the context needed for long analytical journeys. This workspace-first design keeps history, permissions, and intermediate outputs in one place—something harder to manage when analytics happen inside generic chatbots.

Interested teams can explore the GitHub repository or view a demo to see how the pieces fit together.