Agent-built dashboard
An agent-built dashboard is a dashboard that an AI agent authors directly, through an API or an MCP server, producing the same versioned and reviewable document a person would write by hand.
The distinction matters because many “AI dashboard” features produce something else: a chat answer, a chart image, or a one-off HTML page. Those look finished but can’t be updated, reviewed, or reconnected to live data. An agent-built dashboard is a real dashboard that happens to have an agent as its author.
How it works in dvt
In dvt, an agent is any non-human caller: Claude Code in a terminal, a scheduled job, or a CI script. An agent connects to the dvt MCP server, reads your schema and your team’s skills, writes the SQL for each panel, and constructs a dashboard spec. It validates the spec, runs a preview to see the render, and only then applies it.
The result is a live dashboard that runs its queries in your warehouse at view time. Every version records whether a human, an agent, or the system made it, and, for API-key agents, which key it acted through. API safeguards built for agents keep retries and parallel work safe: a repeated request with the same idempotency key returns the original result instead of creating a duplicate, and an update made against a stale version is rejected instead of overwriting someone else’s change. The quickstart builds one from a clean workspace in four prompts.