I build WhatsApp agents and the backends behind them. I am posting here
because the part that most often stalls an n8n team is the part I have
already been through several times.
Where I am useful:
- Meta Cloud API app review, taken end to end including webhook
verification. This is the step that delays a WhatsApp go-live by weeks,
and it is not something a workflow node solves for you. - The 24-hour reply window, handled as an explicit state constraint
rather than an afterthought. - Webhook duplicate suppression and message debounce. Real users send
three messages in a row, and a naive handler answers three times. - Human handoff with locking, so the agent and a human operator cannot
both reply on the same thread. - Conversations in English, Hinglish and Gujarati in production.
Being straight about the fit: I do not build inside n8n. My work is
TypeScript and Node on Cloudflare Workers, with Supabase, Postgres and
the Google Workspace APIs. If you are running n8n, I fit as the WhatsApp
and integration layer beside it, reachable over webhooks and HTTP
requests. If you want the whole thing built as n8n workflows, I am not
your person, and I would rather say so now than three days into a
project.
Public code: GitHub - logiclovingmind/wa-agent-platform: Multi-tenant WhatsApp AI agent platform — Cloudflare Workers, Durable Objects, Supabase Postgres with row-level security. Onboarding a client is a DB insert, not a deploy. · GitHub
A multi-tenant WhatsApp agent platform. 17-table Postgres schema with row
level security on every table, 45 forward-only migrations, and 283 tests
running against the real Cloudflare Workers runtime.
Live agent: whatsapp-demo-agent.vercel.app
One architecture adapted across 10 verticals, where a new deployment needs
credentials and configuration only.
More: logiclovingmind.com
Available now. I am on IST but flexible. Agencies are welcome and I am
happy to deliver white label under your own brand. Send me a real brief
and I will build a working demo before you commit to anything.