Skip to main content

Welcome to Vibedasher

You build dashboards in your own agent harness — Claude Code, Codex, Gemini — and Vibedasher maintains the data platform they run on. That's the split. The dashboard is code: React and SQL your agent writes locally, talking to Vibedasher through the SDK + MCP. Everything underneath the dashboard is stateful, compounding work that doesn't belong in a one-off agent session — connections, datasets, pipelines, refreshes, governance, the query engine, hosting — and that half lives in Vibedasher, maintained and governed.

The loop is: your agent builds against the platform, publishes to it, and Vibedasher hosts, governs, and serves the result — a stable URL, gated by login and by row, that keeps answering after the agent session is over.

Quickstart — point your agent at Vibedasher​

The primary way in is the MCP server. Install it, sign in from your browser, and register it with your agent:

pip install 'vibedasher-mcp>=0.6.0'
vibedasher login
claude mcp add vibedasher --env VIBEDASHER_REGION=eu-central-1 -- vibedasher-mcp

vibedasher login opens your browser, you approve the session, and the refresh token is stored in your OS keyring (a 0600 file if the machine has no keyring). Nothing to mint, copy, paste or rotate. vibedasher whoami shows who is signed in; vibedasher logout ends it.

Or in any MCP client config (Claude Code, Cursor, …) — no key in the env block:

{
"mcpServers": {
"vibedasher": {
"command": "vibedasher-mcp",
"env": { "VIBEDASHER_REGION": "eu-central-1" }
}
}
}

From there your agent can upload data, run SQL, create and build dashboards, publish them, and manage who sees what — see Build with your agent for the full walkthrough. Running on CI or a headless box where no browser can open? That lane uses an API key instead — see Headless and CI.

The platform Vibedasher maintains​

Your agent writes the dashboard; Vibedasher maintains everything the dashboard depends on:

  • Connections — 19 connector types available in the app today (PostgreSQL, MySQL, Snowflake, BigQuery, Redshift, Salesforce, and more) plus CSV and Excel upload, set up once and queried by every surface — MCP tools, SDKs, the app. The full list, including what is not available yet, is on the connections page.
  • Datasets & pipelines — ingested, schema-introspected datasets with a stable SQL alias; dbt projects for transformation; refreshes run by our cron, or triggered by you or your agent — every schedule evaluated in UTC.
  • The query engine — typed SQL over your datasets, with result caching, exposed identically to agents (MCP), code (SDKs), and published dashboards.
  • Governance — real accounts, roles, groups, and allow/deny access rights on folders, inherited by the dashboards and datasets inside them — plus per-viewer row-level filters on embedded dashboards, enforced server-side from a signed token. Not a shared password. See Who can see what for what the model covers and what it does not.
  • Hosting & serving — a published dashboard gets a stable URL that keeps answering after the agent session is over; public, behind your login, or embedded in your own product. What we host is a client-rendered bundle: React running in the visitor's browser over data served by SQL, with none of your server code in the request path — so a server-rendered framework (Next.js, Remix, SvelteKit, Astro, Nuxt) cannot be hosted as it stands. That stack is routed rather than refused: keep your own server and pull this account's data into it with the SDK, or eject the dashboard and host it yourself. Worth knowing before you build, not after.

And the dashboard code itself stays yours — it's plain React and SQL you can eject and host elsewhere, any time. A public, published dashboard ejects onto a credential-free data lane: no key, no token, no backend of your own.

Where to go​

Vibedasher Cloud or Vibedasher Server​

Same architecture, two ways to run it:

  • Vibedasher Cloud — fully managed at cloud.vibedasher.com, priced by usage, not seats. Create an account, run vibedasher login, point your agent at it.
  • Vibedasher Server — self-hosted in your own VPC, with your own LLM if you use the in-app assistant. Your data never leaves your environment.
note

Vibedasher is under active development and these docs grow alongside it — if something is missing or unclear, reach out via the contact page.