Ask your data. Keep the dashboard.
Tell the coding agent already signed in on your machine what you need. It creates an HTML dashboard in your Valv workspace. Open the file later and it runs its saved database queries again, with no BI server to deploy.
Already have Valv? See how live dashboards work.
One prompt becomes a file.
Ask for the metrics and breakdowns you need. The agent tests its data calls, writes one HTML file, and opens it beside the conversation.
Revenue by month
The file asks again.
Opening or refreshing the dashboard runs its saved structured queries against the active workspace. That refresh does not start another AI model turn.
const rows = await valv.query("DATABASE_ID", {
from: "orders",
select: { revenue: { sum: "amount" } }
});Structured query · current rows
From a database connection to a dashboard you can reopen.
The setup stays small because Valv uses your database connection and the coding agent account you already have.
- 1
Connect a database.
Add PostgreSQL, MySQL or MariaDB, SQLite, ClickHouse, or experimental MongoDB. Valv encrypts saved passwords through the OS keychain and runs queries from the desktop app.
- 2
Review AI access.
All columns in the active schemas start visible to the agent. Hide any table or column it should not discover before asking the first question.
- 3
Ask for the analysis.
Use Claude Code, Codex, GitHub Copilot, Cursor, Grok, or OpenCode through its existing login. The agent can test database queries and turn the result into a dashboard file.
- 4
Open it again when the data changes.
The HTML file stays in the workspace and appears under Artifacts. Each open or manual refresh runs the saved calls again and renders the latest rows.
Local execution is not the same as no data leaving.
Valv does not send database credentials to the agent or route queries through a hosted BI service. The AI provider still processes the data needed to author and discuss the analysis.
While the agent builds.
Valv launches the agent installed on your machine. Prompts, allowed schema details, and query results the agent needs are processed under that provider account. Credentials remain inside Valv.
Whenever the file opens.
The saved dashboard queries through Valv's desktop bridge. Its frame cannot make its own network data requests. It may load Chart.js from jsDelivr, the one script origin allowed by the dashboard policy.
When you choose to share.
In a published git workspace, the share control syncs the selected file through your git remote. Shared datasource configs contain no passwords. Each teammate supplies credentials on their machine.
Set access before the prompt.
The active schemas start available to the agent. Hide a whole table or individual columns, then ask the agent to build. Your own table and SQL views keep their normal access.
Changing AI access later does not rewrite a dashboard that already contains a saved query. Review or remove old files when access rules change. The database user's permissions remain the final boundary.
AI access
Share the file, not a screenshot.
The share control appears in git-backed workspaces. After you publish the workspace, Valv commits and pushes selected files. Teammates open the same dashboard against the connection configured on their machine.
Files start private. Chat histories and database passwords are not part of the shared git surface.
MRR
$48,400
vs last month
+12%
Free for personal analysis. One payment for commercial work.
Personal projects, learning, and unpaid open-source work are free. During the beta, commercial use costs $99 per person once and includes lifetime updates. You keep using and paying your AI provider directly.
AI business intelligence, answered
What is AI business intelligence?
AI business intelligence uses an AI agent to answer data questions and build analysis. In Valv, your chosen coding agent can create an HTML dashboard in the workspace. The saved dashboard runs its structured database queries again whenever you open or refresh it.
How is this different from Tableau, Looker, or Power BI?
Valv is a desktop database client, not a hosted BI platform. You connect a database, ask the agent for a dashboard, and keep the result as an HTML file in the workspace. There is no BI server or semantic-layer project to deploy before the first question.
Does my data leave my machine?
Database credentials stay in OS-protected storage, and Valv runs database queries from your computer. While the agent builds or discusses an analysis, the prompt, schema, and query results it needs are processed through the provider account used by your local agent. Reopening a saved dashboard queries through Valv without starting another model turn.
Which databases work with AI dashboards?
PostgreSQL, MySQL and MariaDB, SQLite, and ClickHouse are supported. MongoDB support is experimental. One workspace can contain multiple databases, and a dashboard can query the connected data through Valv.
Do I need to know SQL?
No. Ask in plain language. Valv's database tools use structured query objects, and a live dashboard saves those objects in its HTML. You can inspect or edit the file, while Valv's separate SQL editor is available when you want to write raw SQL yourself.
What can the agent access by default?
A new database starts with every column in its active schemas visible to the agent. Review AI access before your first question and hide any table or column the agent should not discover. This controls agent discovery and queries. It does not change the database user's permissions or remove queries already saved in a dashboard file.
How much does Valv cost?
Personal projects, learning, and unpaid open-source work are free. Commercial work needs a license per person. During the beta, Pro is a one-time $99 per seat and includes lifetime commercial use and updates. Your AI provider account is separate.
Ask for your first dashboard.
Free for personal use, every feature included. Commercial use needs a seat at beta pricing. See pricing.