Travel Agency WhatsApp Booking Bot — Open-Source Template

Placeholder case study — full write-up coming soon. The shape of the project is below.

The problem

After building the same WhatsApp booking bot for the third travel agency, I realised the core flow was identical every time: capture the package, dates, pax count, budget, push to a sheet, hand off to a human. Every agency wanted it, every agency was being quoted £6–10k by "agencies" for what is essentially one n8n workflow.

What I built

An open-source n8n template + setup guide that any travel agency (or freelancer) can self-host on a $5 VPS.

What's in the box

• n8n workflow JSON for the full conversation flow
• WhatsApp Business API setup guide (Meta Cloud API + verified business + template messages)
• Chatwoot integration for human handoff
• Postgres schema for conversation memory + lead storage
• Docker compose file so it all comes up with one command

Conversation design

• Intent detection: pricing, availability, visa, payment, change/cancel
• Slot filling: destination, dates, pax (adults / children / infants), budget range
• Fallback to a human agent the moment confidence drops or the user says "agent"
• Multilingual: English + Urdu, with Arabic auto-translation experiment

The hardest part

State management without a database for v1. The first version tried to hold state in n8n itself between webhook calls. That works for 5 conversations and breaks at 50. The fix was adding Postgres for conversation memory keyed by phone number. The template ships with that already wired up so no one else burns a weekend on the same mistake.

Outcome

• Reusable foundation for every new travel client — saves 3–4 days per project
• Lower selling price for agencies who can't afford a custom build
• Public repo as a credibility asset for inbound leads

What I'd do differently

I'd ship a deploy-to-VPS button (1-click DigitalOcean / Railway / Render setup) so agencies who don't have a developer can still self-host without copying a docker-compose by hand.