Your entire data stack.
One open-source platform.
ETL, dbt orchestration, a SQL client, dashboards, and an app builder — five layers that build on each other, run by an AI agent that knows your schema. Self-hostable, MIT-licensed, no credit card.
This is a live miniature, not a screenshot — click through the sidebar, tabs, chats, and result views. The real thing just also runs your queries.
Five tools you can delete
Each module builds on the one before it. Start with the layer that hurts most — the rest is already there when you need it.
ETL without the pipeline vendor
Your ETL layer, like Airbyte or Fivetran — connect Stripe, PostHog, Close, or any REST API once, then let Flows sync them into your warehouse on a schedule, with run history per flow. No pipeline vendor, no per-connector pricing surprises.
Explore ETL Connectors & FlowsThe AI-native SQL client
Your SQL client, like DataGrip but in the browser — Run is ⌘/Ctrl+Enter, results flip between table, JSON, and chart, and the agent writes SQL against your real schema. Consoles and connections are shared: My Consoles for you, Workspace for the team.
Explore SQL Client| # | plan | month | mrr |
|---|---|---|---|
| 1 | Scale | 2026-06 | 91,860 |
| 2 | Team | 2026-06 | 74,220 |
| 3 | Starter | 2026-06 | 47,930 |
| 4 | Hobby | 2026-06 | 21,510 |
| 5 | Scale | 2026-05 | 84,130 |
| 6 | Team | 2026-05 | 71,890 |
Dashboards the agent builds for you
Your BI layer, like Tableau or Looker Studio — except you describe the dashboard and the agent assembles the tiles from your saved queries and dbt models. Flip between View and Edit, or drop into the code behind any widget.
Explore DashboardsFull React apps on your data layer
Your app builder, like Lovable or Retool — but the data layer is already bound: models, saved queries, auth. Prompt your way to a full React app, iterate in chat next to a live preview, then hit Publish.
Explore AppsOne agent runs through all five layers. It syncs the data, models it, queries it, charts it, and ships it — and its per-workspace memory means it gets better at your stack the more you use it.
Loved by people who deleted their stack
“We replaced a pipeline vendor, an orchestrator, a BI tool, and two desktop licenses with one docker-compose file. The council was impressed.”
“The agent fixed our nightly dbt run before I'd finished my caf. I felt a great disturbance in the on-call rotation.”
“I asked for an account health app in plain English. It shipped one wired to our real models. These are the dashboards I was looking for.”
Testimonials transmitted from a galaxy far, far away.
Open source. Self-hostable. Yours.
The whole platform is MIT-licensed. Run it on your own infrastructure, keep your data where it lives, and never talk to a sales rep unless you want to. No credit card, no trial clock, no feature gates.
Frequently asked questions
What is Mako?
Mako is an open-source AI data platform. It covers the whole loop in one tool: Connectors and Flows pull your SaaS data into your warehouse, Transforms runs your dbt project with an agent on call, Consoles query every database you use, Dashboards chart the results, and Apps turn them into real React tools.
Do I have to adopt all five modules?
No. Most teams arrive for one layer — usually Consoles or the connectors — and grow into the rest. The modules share connections, schema knowledge, and the agent's memory, so each one you add starts smarter than a standalone tool would.
How does this replace Airbyte or Fivetran?
Mako ships connectors for Stripe, PostHog, Close, and generic REST APIs, with scheduled Flows into your warehouse and run history per sync. If your sources are covered, you don't need a separate extraction vendor or its pricing model — and the data lands somewhere you can immediately query.
What does the dbt orchestrator actually do?
It runs and schedules your existing dbt project, like dbt Cloud — and then goes further: the agent writes new models on request, diagnoses failed runs against your live warehouse schema, and proposes fixes as reviewable diffs.
How is this different from Lovable or Replit?
Prompt-to-app builders start from a blank database. Mako's app builder starts from your data layer: the synced sources, dbt models, and saved queries underneath it. You describe the app; it's already wired to numbers that are real and fresh.
Can I use Mako from Claude Code or Cursor?
Yes — Mako is also an MCP server. Create a workspace API key and point Claude Code, Cursor, or Codex at /api/mcp: your own agent can then explore your schema, run read-only queries, and build Mako apps using your existing AI subscription.
Is it really free?
Mako is MIT-licensed and self-hostable with all features included — no credit card, no trial. If you'd rather not run it yourself, there's a hosted version at app.mako.ai.
Where does my data go?
When you self-host, your data stays on your infrastructure. Mako connects to your databases directly; you own the deployment end to end.
Start deleting tools.
Connect a database and ask it something. The rest of the platform is there when you need it.