dbt Labs has released dbt Charts, an open-source language for building dashboards inside a dbt project, announced Monday at dbt Summit 2026 at dbtcharts.com. Dave Fowler, general manager and head of the project, and designer RJ Andrews introduced it.

Fowler framed the problem as a split between two flawed approaches. AI tools can now vibe-code a dashboard from a prompt, but each result mixes JavaScript, Python, CSS and charting libraries differently, making consistency hard and tracing a chart back to its underlying query harder still. Traditional BI tools, meanwhile, are UI-first products with AI bolted on afterward, which limits how much an agent can change.

dbt Charts uses a simple YAML syntax built around variables, queries, charts and rows for layout - a kind of markdown for dashboards, Fowler said. It adds a charts folder beside a project's models folder, so renaming a column or updating a model runs through the same git tests and deploys in the same branch as everything else.

The language is open source under Apache 2.0, the same license as dbt, and runs locally through a CLI called dct. A hosted product at dbtcharts.com adds permissioning, sharing, a visual editor, version history and a conversational AI interface.

Andrews, a historian of data graphics, described the design defaults: direct labels instead of legends, decimal-aligned support tables so numbers scan cleanly down a column, and axis labels placed on whichever side keeps a reader's eye on the most recent data. The aim is for a dashboard built entirely on defaults, with no custom styling, to already look worth reading. In a live demo, Fowler edited chart titles, colors and markdown text on a dashboard, then reverted the page to an earlier version using its built-in git history.

Source: dbt Labs Blog