I first launched ToolJet here in 2021 as a one-person project. It blew up really well & got 1k stars on GitHub (https://github.com/ToolJet/ToolJet/) in around 8 hours. Back then ToolJet was basically a frontend builder that could connect to different data sources.

Since then, we kept expanding:

- Added a workflow automation tool so builders could orchestrate background jobs.

- Added a built-in no-code database so builders didn’t need to spin up a new db.

- Eventually grew into a full-stack platform for internal tools.

- And other obvious things like tons of smaller features & integrations.

But last year we kind of messed up. We kept adding features, the frontend architecture couldn’t keep up, and stability/performance issues showed up once apps got complex (ie hundreds of UI components in a single page of an app). So we stopped, rebuilt the architecture (ToolJet v3 in November), and cleaned up a lot of tech debt. That gave us a solid foundation - and also made us realize it was the right moment to go AI-native.

We analyzed how our users actually built apps: 80% of the time on repetitive setup (forms, tables, CRUD), 15% on integration glue code, 5% on actual business logic. Traditional low-code tried to eliminate code entirely. We're eliminating the wrong code - the boring 95% - while keeping full control for the 5% that matters.

Instead of "prompt-to-code," ToolJet AI tries to copy how an engineering team functions (yeah, a bit opinionated way) - but with AI agents:

- PM agent → turns your prompt into a PRD.

- Design agent → generated the the UI using our pre-built components and custom components.

- DB agent → builds the schema.

- Full-stack agent → wires it all up with queries, event handlers, and code.

At each step, builders can review/edit, stop AI generation, or switch into the visual builder. Generated apps aren’t locked in - you can keep tweaking with prompts, drag-and-drop, or extend with custom code.

Why this works:

We know "AI builds apps" is overhyped right now. The difference: we're not generating raw code - we're configuring battle-tested components. Think Terraform for internal tools, not Claude/GPT writing React.

That means:

- Fewer tokens → lower cost.

- Deterministic & Faster outputs → fewer errors.

- More reliability → production-ready apps.

Basically, AI is filling in blueprints.

ToolJet AI is a closed-source fork of the open-source community edition, which will continue to be actively maintained. All the core platform changes (like the v3 rebuild and stability/performance work) are committed upstream. The AI features sit on top, but OSS remains the foundation.

Thanks for reading - and thanks again to HN for being part of ToolJet’s journey since the very beginning.