Why build your dashboards in Tableau when Claude Code builds them now?

The dashboard your AI builds needs a home — a URL, a login, row-level permissions. That's us.

No credit card · Works with the agent you already use · Open beta

ONE PROMPT

Watch Claude Code build the whole dashboard — then publish it.

Two acts, one real session, nobody steering. Act one: it pulls the vibedasher playbook, extracts the data through the MCP, writes the app, builds its own browser harness and checks every headline number against the warehouse to the cent. Act two: it publishes the dashboard to vibedasher through the same MCP — a URL, behind your login, live against the dataset — and you see it working: tooltips, metric switching, cross-filtering, dark mode.

claude code · vibedasher mcp · two acts · 71 s

Real sessions, cut for speed. Thirteen self-corrections in the build, all on camera — the agent catching its own mistakes is the part no one can fake.

The prompt — the entire input
Build me a sales performance dashboard from the Superstore data in my vibedasher account. I want it running locally in the browser — React, my own repo, not inside your app.

It needs to be genuinely good, not a toy: every number shown against the same period last year with the change, clicking a category or region should filter the whole dashboard, hovering a category should highlight it everywhere, a US map by state, and rich tooltips. It should look right on my phone too, and in dark mode.

Work in this folder. Don't ask me questions — use your judgement and just build it.

Your agent took over the building. These jobs remain.

Exploring the data, writing the SQL, drawing the charts, iterating on layout — that all moved into Claude Code, Codex, Gemini. What's left is the data platform every dashboard stands on — and that's what we maintain:

Publish

A real URL that keeps answering after the agent is gone — not a file to email around. Your agent pushes it through the MCP or SDK; we host and serve it.

Govern

Real accounts, groups, and allow/deny access rights on every folder, down to the row. Not a shared password.

Share

A link, a login-gated view, or an embed in your own product with a single iframe.

Your agent builds the dashboard, in your harness. We keep the platform under it running — connections, datasets, dbt pipelines, refreshes, permissions, the query engine — so it's still standing next quarter.

From your agent to a shared, permissioned dashboard

Three steps. The agent your team already runs does the building; we do everything that has to still be there next quarter.

1

Point your agent at your data

Install the vibedasher MCP server; your harness reads schemas and runs SQL against your warehouse.

2

Ask for what you want

Plain language, in the terminal or editor your team already uses. The agent writes the React + SQL.

3

It publishes into the environment

A URL, a login, a folder, access rights, and an embed snippet — hosted and governed, ready to share.

Run it as Vibedasher Cloud or Vibedasher Server.

Vibedasher Cloud

We run it. Sign up and your agent publishes in minutes — nothing to install, nothing to operate.

Vibedasher Server

Self-hosted in your own VPC, with your own LLM — air-gapped if you need it. Set up with our team.

Same code. Same architecture. Your call.

Everything a BI server does — sold on its own

Deliberately familiar. A BI manager expects all of it; that's the point.

Host & serve

A URL that keeps answering after the agent finishes. Public or behind your login; embeddable in your product.

Govern

Real accounts, groups, and allow/deny access rights on every folder — row-level. Not a shared password.

Connect

19 connectors — Postgres, Snowflake, BigQuery, Salesforce… — plus CSV/Excel/Parquet upload, datasets, dbt pipelines, triggered refreshes, and a typed query engine.

Run it your way

Vibedasher Cloud, usage-based — or Vibedasher Server in your own VPC. Same code either way; see above.

Tableau · Power BI · Looker

The lock-in was the file format.

For twenty years BI lock-in was simple: your dashboards lived in a format only your vendor's builder could write — a Tableau workbook, a Power BI report and semantic model, a LookML model — and only that vendor's server could host. Leaving meant rebuilding everything, so nobody left.

Agents broke the format. A dashboard your agent builds is React and SQL — a format everyone owns and no one can hold hostage. The old servers can't host it; it doesn't need them.

Vibedasher is a server for the unlocked artifact. And it can't re-lock you: eject any dashboard as plain React you own, any time. If you ever leave, you take the code with you.

Tableau

Its server hosts

A Tableau workbook, authored in its own builder. No slot to host a standalone React + SQL app.

Power BI

Its server hosts

A report bound to a semantic model, published from Power BI Desktop. No slot to host a standalone React + SQL app.

Looker

Its server hosts

A LookML model compiled to Looks and dashboards. No slot to host a standalone React + SQL app.

vibedasher

It hosts

Your agent's React + SQL — hosted, governed, and served as a first-class, permissioned dashboard. Ejectable any time.

See a live ejected dashboard.

Pricing

Pay for the engine, not for seats.

One idea, both editions: you pay for what the engine runs — never per-seat, never per-view. Sharing a dashboard costs nothing.

Vibedasher Cloud

Usage-based, metered on engine work.

  • Metered on the queries and compute your dashboards actually run — nothing else.
  • Never per-seat or per-view — invite the whole company, a viewer costs nothing.
  • Start free — no credit card.

Vibedasher Server

Self-hosted; priced with our team.

  • Runs in your own VPC, on your own infrastructure.
  • Your own LLM — you pay your model provider directly; we never resell inference.
  • Set up and sized with our team — no seat count in the contract.

Full pricing →

Keep the dashboards. Drop the licence.