dbt Charts: Open-Source YAML Language for AI-Built Dashboards
Original: Charts built for Chat
Why This Matters
If AI agents become the primary dashboard builder, a code-native, auditable chart format becomes table stakes for data governance.
dbt Charts, a declarative YAML language for dashboards, is now open source. Announced September 14, 2026 by Dave Fowler, it lets AI agents generate auditable, governed dashboards in a single YAML file — avoiding the sprawl of HTML, CSS, JS, and React files that agent-built reports typically produce.
As AI agents make self-serve analytics real, a new problem has emerged: agents asked to build a report tend to output a tangle of HTML, CSS, JavaScript, and framework files that are hard to audit and expensive to update token-by-token. BI tools tried adding copilots to their UIs, but that limits agents to whatever the interface exposes.
dbt Charts is positioned as a third path. It moves charts out of the BI tool and into a structured YAML file where SQL declares what data to fetch, YAML declares how to display it, Markdown carries prose, and Jinja (familiar from dbt) handles variables and macros. One file, one dashboard, fully version-controlled and readable.
The launch fits into a broader narrative the team calls BI's second unbundling. The modern data stack (2015–2022) already pulled warehousing, ingestion, and transformation out of monolithic BI platforms. What remained — visualization, semantic definitions, IAM, hosting — is now the BI tool people use daily. The argument here is that chart creation is next to leave, because agents are fluent in code and Git but clumsy in someone else's UI.
The project is open source on GitHub. A beta platform and pricing are available at dbtcharts.com.