Build exactly the view you want. Know exactly how it got there.
dvt lets anyone build rich, working dashboards by describing them to the AI they already use, faster than a traditional BI tool. Every dashboard carries its full history: who built it, when, from what data, and why.
Describe it. See it. Change it.
Tell your AI what you want to see and dvt builds it: the charts, the layout, the numbers behind them. Ask for a different cut of the data, a new page, a chart moved or restyled, and it changes in seconds. You never have to drag and drop to get what you want, and there is no chart type you cannot adjust.
dvt works inside the AI tools your team already runs: Claude Code, Snowflake Cortex, or any assistant that speaks MCP. There is no separate AI to adopt. dvt is the part that turns what your assistant understands about your data into something the business can look at.
Everything the AI can do, a person can do too: open any dashboard and edit every property by hand.
Every number has a paper trail.
When anyone can build a dashboard, the question becomes trust. dvt answers it by recording, for every dashboard, page and chart: who built it, whether that was a person or an AI agent, when, what data it reads, what they assumed, the credential an API or agent write came through, and any reason left with the change.
That history is not a log you export. It is something you can ask about, in plain language, through the same AI that built the dashboard.
Ask dvt. It finds every dashboard that shows churn, shows how each one defines it, who built each definition, what it assumed, and which one the team actually looks at. What used to be a week of meetings is a side-by-side comparison in minutes.
Every edit is a version with who made it and when, plus a reason when the editor leaves one. Compare any two versions, see exactly what moved, and restore an earlier one in one step.
Every chart records the query behind it, and dvt traces that back to the tables it reads. Every dashboard records what it was built to answer and for whom. Nobody has to write documentation after the fact. It is written as the dashboard is built.
Who uses dvt
Data leaders and BI admins
You are accountable for the numbers the business runs on. dvt gives you a complete record of who built every dashboard, what it reads, and how it changed, plus permissions by role and one house style set once for every dashboard.
“I can finally answer ‘where did this number come from?’ without a meeting.”
Analysts and business teams
You know what you want to see and you use an AI assistant every day. Describe the dashboard, get it back working, and adjust anything by hand.
“I asked for a revenue dashboard and it just worked.”
Analytics engineers
You already model data in dbt and want the same control over what people see. dvt dashboards are specs you can generate, version, review and put in git when it makes sense.
“Dashboards I can review like code.”
Where dvt fits
Your warehouse already has good tools for loading data (dlt) and modelling it (dbt). The last step, turning it into something people look at, is still a point-and-click product that AI cannot drive and admins cannot audit. dvt is that last step, built for the way teams work now. It does not replace your warehouse or your metric definitions. It sits on top of them and keeps the record.
The complete modern data stack
Every dashboard in your company built by whoever needed it, in minutes, with a history anyone can read.
dvt dashboards are a documented JSON spec. Any AI agent can write one, and a spec that stays in the dvt Core profile is portable across any compliant renderer. That is the bet: when building is this easy, the record of how things were built becomes the most valuable thing you own.
See the spec →Follow the build
dvt is in founding research. Leave your email and we'll reach out when there's something to see — no newsletters, no noise.