automate workflow

Inspiration

Most no-code automation tools ask you to click through 10 dropdowns before you've built anything. I wanted the opposite — you describe an automation in one sentence and it appears on the canvas, editable and ready to run.

What it does

Provide ai chat panel to create workflows, which are visible node by node on the canvas with live execution, with history logs

How we built it

Frontend: Next.js 16 (App Router) + React 19 + TypeScript + Tailwind v4. React Flow for the canvas, GSAP for the 3D ring hero on auth pages, Zustand for canvas state, TanStack Query for server state. AI layer: Anthropic Claude + OpenAI SDK behind a single session-scoped chat endpoint. The model output is validated by a Zod schema before ever hitting the canvas — no free-form JSON reaching state. Backend: Node.js + Express + MongoDB. JWT auth, tenant-scoped everything, usage metering (trial limits on AI calls + test runs). Worker: BullMQ + Redis for job queue, embedded into the server process for a single-container Railway free-tier deploy. Real-time run events streamed to the client over Socket.IO. Email: Brevo HTTPS API primary, Resend fallback, SMTP fallback for local dev. Deploy: Vercel (client), Railway (server + Redis + Mongo). Theming: CSS-variable-driven light/dark with a pre-hydration script to prevent flash.

Challenges we ran into

Resend free tier only delivers to the account owner's email. Fine for testing, broken for a real demo.

Accomplishments that we're proud of

End-to-end AI generation that actually works with real time execution streaming.

What we learned

When prod times out and local works, debug the network, not the code.

What's next for automate workflow

Slack step activation

Built With

• express.js
• gsap
• mern
• mongodb
• node.js
• react
• redis
• socket.io
• typescript
• zod
• zustand