Tech stack
The tools we build with, and why
No tool is the answer on its own. Here is what we use, what each one is good for, and how we decide.
AI models
We pick the model per task and design so you can switch providers later without a rebuild.
- Anthropic Claude
- Long documents, careful writing, coding help and agent work
- OpenAI GPT models
- General assistants, structured data extraction, vision and voice
- Google Gemini
- Teams already on Google Workspace, very long inputs
- Open models (Llama, Mistral, Qwen)
- Running on your own server when data cannot leave it
Agents and AI frameworks
For assistants that take several steps or use your tools, not just answer a question.
- OpenAI Agents SDK
- Tool-using agents with handoffs and guardrails
- LangGraph
- Multi-step agents that need state, retries and human approval
- Vercel AI SDK
- Streaming AI features inside Next.js and React apps
- Model Context Protocol (MCP)
- Connecting assistants like Claude to your own tools and data
Knowledge and data (RAG)
Search over your own documents with answers that cite the source.
- PostgreSQL + pgvector
- Our default: documents, embeddings and app data in one database
- Supabase
- Managed Postgres with auth and storage for smaller teams
- Qdrant or Pinecone
- Large document sets that need a dedicated vector store
- NotebookLM
- Quick, no-code research assistants for students and small teams
Automation platforms
The glue between your apps. Chosen by who will maintain it and where your data must live.
- n8n
- Our first choice: visual workflows, self-hosted when you need control over data
- Make
- Teams that want a hosted visual tool with many ready connectors
- Zapier
- Simple flows for people who already use it
- Custom workers
- High-volume or business-critical flows that need code and tests
Backend and APIs
When a workflow tool is not enough: webhooks, queues and integrations written as code.
- Python + FastAPI
- APIs, data processing and AI services
- Node.js + TypeScript
- Integrations and services next to a JavaScript product
- PostgreSQL
- The source of truth for anything that must not be lost
- Redis and managed queues
- Retries, rate limits and background jobs
Web apps and interfaces
Dashboards, internal tools, portfolios and customer-facing AI features.
- Next.js + React
- Websites, dashboards and internal tools
- Tailwind CSS
- Fast, consistent interfaces that are easy to change
- React Native
- Small mobile companions to an existing system
Cloud and DevOps
Hosting, deployment and monitoring, in your own accounts.
- AWS and Google Cloud
- Production workloads and self-hosted automation
- Vercel
- Next.js sites and apps
- Docker
- The same setup on a laptop, a server or the cloud
- GitHub Actions
- Tests, deployments and scheduled jobs
- Sentry
- Error alerts with enough context to fix them
Business tools we connect
We work with the tools you already use. These are the ones we connect most often.
- WhatsApp Business API
- Lead replies, order updates, reminders and support
- Zoho CRM, HubSpot, Salesforce
- Leads, deals and follow-ups
- Tally, Zoho Books, QuickBooks
- Invoices, payments and reports
- Razorpay, Stripe, UPI links
- Payment links and payment status
- Shopify, WooCommerce, Shiprocket
- Orders, dispatch and delivery tracking
- Google Workspace, Microsoft 365
- Email, calendar, sheets and documents
- Slack, Notion, Airtable
- Team updates, wikis and lightweight databases
- Cal.com, Google Calendar
- Booking calls and interviews
How we choose
01
What you already pay for comes first
If your CRM or sheet can do the job, we use it before adding anything new.
02
You can maintain it without us
We choose tools your team can open and understand, and we document every one.
03
Your data stays where it should
Sensitive data can stay on your own server with self-hosted n8n and open models.
04
Running cost in writing
Before we build, you get the expected monthly cost of every tool and AI call.

