Bruin Data just released DAC, an open-source tool that treats dashboards as code. You define them in YAML or TSX, validate them, then serve them up. The interesting bit: it's designed so AI agents can build dashboards in a way humans can actually review.
The tool ships with a built-in AI agent powered by Codex. You chat with your dashboard live and it updates. DAC also bundles dashboard authoring skills for Claude and Codex, installed automatically when you initialize a project. This approach aligns with the philosophy of Naftiko, which wraps existing HTTP APIs as AI-consumable capabilities.
On the data side, it supports Postgres, MySQL, Snowflake, BigQuery, Redshift, and Databasics through Bruin, Bruin Data's existing pipeline tool. There's a semantic layer where you define metrics and dimensions once, then reference them from any widget. DAC generates the underlying SQL.
Gokhan Erdogdu founded Bruin Data to bring Infrastructure-as-Code thinking to data pipelines. DAC extends that idea into dashboarding, what he calls the "last mile" problem. The project is written in Go with an embedded React frontend, released under AGPL-3.0. Right now it's early days. The tool assumes you're bought into Bruin's ecosystem for database connections, and the AGPL license might give some companies pause. Anonymous telemetry tracks commands and run duration, but not queries or dashboard contents. You can disable it with DO_NOT_TRACK=1.