Most charting tools assume the field you want to plot already exists as a column. Data Formulator bets the opposite: the view you actually want usually needs a derived measure — a ratio, a rank, a reshaped category — that isn't in your table yet. Its core move is letting you bind the fields you have by drag-and-drop while naming the ones you don't in plain language, then having an AI agent write the transformation that materializes them.
What Sets It Apart
- Concept-driven binding over code-then-plot. You declare the chart you want and the missing fields it needs; the agent generates the transformation, so iteration happens at the level of intent rather than pandas snippets.
- A Data Agent with thread memory. Analysis paths are saved as branchable threads you can revisit and compare, which means exploratory dead-ends don't erase the trail back to a working state.
- Ingestion from non-tabular sources. It pulls structured data out of screenshots, websites, Excel, and free text, so the messy first-mile of getting numbers into a chart is part of the loop instead of a manual prep step.
- Breadth of output. 30+ chart types — including streamgraphs, candlesticks, radar, and maps — with natural-language styling, covering cases most quick-viz tools punt on.
Who It's For
Great fit if you live in exploratory data analysis and keep hitting the wall where the chart you imagine requires a field you'd have to engineer first — Data Formulator collapses that gap. Look elsewhere if you need pixel-locked, governed dashboards on a fixed schema, or if sending data to an external LLM is a non-starter; the iterative, agent-mediated workflow trades reproducibility and control for speed of discovery.
