The tech stack behind our AI and SaaS builds

At a glance

Clients often ask what we build with. This is the stack our AI engineer, Nauman N., uses across the AI and SaaS projects on our Work page, grouped by the job each part does.

The app people use

Next.js, React and TypeScript for the web app, with FastAPI (Python) or Node.js behind it. PostgreSQL or Supabase store the data. Stripe handles payments and subscriptions, as in the pet care booking marketplace and the custom GPT community platform.

The AI layer

The Claude API and the OpenAI API handle language tasks, with LangChain and LangGraph running the agents. Retrieval-augmented generation (RAG) with a vector database lets an AI answer from your own documents. In the legal research platform it runs on a self-hosted model, so client documents never reach a third-party API. MCP gives agents a standard way to use tools.

Voice

ElevenLabs, Vapi, Retell AI and Twilio power voice agents and phone calls, such as the AI phone receptionist that books appointments during the call.

Automation and CRM

n8n, Make and Zapier connect apps. GoHighLevel, HubSpot and Airtable hold leads and customer records. The Shipbling project, for example, uses n8n workflows to automate a Shopify store.

Hosting and quality

Docker and AWS for hosting, with automated tests on the larger builds: 93 on the OrbitOS profit dashboard and 243 on the Airtime Media voice agents.

What we look at when choosing

• The tools your team already uses.
• Keeping your data in systems you own.
• The simplest setup that does the job, with agents only where they save real work.

Based on our AI engineer's Upwork profile and portfolio, October 2026.