Microsoft's Data Formulator 0.7 Turns Enterprise Databases Into AI-Guided Insights
Microsoft's open-source Data Formulator 0.7 brings agentic data exploration to enterprise sources — no SQL or coding required
Takeaways
- Data Formulator 0.7 is now stable: a free, MIT-licensed, open-source AI data analysis tool from Microsoft Research with 15.8k GitHub stars.
- New Data Connectors support persistent, governed connections to BigQuery, MySQL, PostgreSQL, Cosmos DB, S3, Azure Blob, Kusto, Superset, and more.
- A unified DataAgent with thread memory tracks every question, chart, and intermediate result across long sessions, with branching support.
- Supports 30+ chart types and a style-refinement agent; works with OpenAI, Anthropic, Azure, or local Ollama models via LiteLLM.
- Install instantly with
uvx data_formulatororpip install data_formulator; also available via Docker and GitHub Codespaces. - Built on peer-reviewed research (arXiv:2408.16119); 0.8 alpha already in pre-release with model routing and enterprise isolation improvements.
Data Formulator, Microsoft Research's open-source AI-powered data analysis tool, just shipped version 0.7 , and it's a meaningful step beyond a charting assistant. What started as a research prototype for turning CSVs into visualizations has grown into a full agentic workspace that connects to enterprise databases, maintains session memory across long analyses, and lets non-technical users explore data through a mix of drag-and-drop and plain English. It has crossed 15,800 GitHub stars and is actively maintained by the Microsoft Research team.
The problem it's actually solving
Enterprise data workflows are often fragmented across storage systems and tools. Before any analysis can begin, someone has to wire up connections, manage permissions, and figure out how to combine data from five different places. Then the actual analysis requires SQL or Python skills that most domain experts don't have. LLM-powered systems have improved visualization authoring, but they don't work well for iterative workflows because they often require analysts to provide, in a single turn, a text-only prompt that fully describes the complex visualization task , which is unrealistic for both users and models.
Data Formulator's answer is a blended interface: you drag fields onto a chart canvas to express structure, and type natural language to express intent. The AI fills in the gap between the two , writing the data transformation code, running it in a sandbox, and rendering the result.
What's new in 0.7
Version 0.7 turns any data into insights in five steps: connect governed data sources, load via agents, explore with the unified DataAgent and Data Thread, refine 30+ chart types with a style-refinement agent, and share as reports. The team describes it internally as a redesign significant enough to warrant a v1.0 label.
The headline additions:
- Data Connectors: Version 0.7 added persistent data connectors to Superset, Kusto, Cosmos DB, MySQL, PostgreSQL, MSSQL, S3, Azure Blob, BigQuery, and others, with SSO, lazy catalog loading, and smart filtering. Platform teams set these up once; everyone else just uses them.