Build a WhatsApp support bot that crawls any website with OpenAI

Get Job Alerts
One email when a teaching platform opens hiring or a new AI training project drops. No weekly filler.

Most small businesses pay for bloated chat platforms that still need constant manual updates. This AI automation crawls your client's website on demand and answers real customer questions over WhatsApp, with zero document uploads or model training required.

What This Automation Does

Automatically crawls any client website, including product pages, FAQs, and policies, so the bot always answers from live content without you touching a file
Handles incoming WhatsApp messages in real time, switching between free-form replies and Meta-compliant templates depending on the conversation window
Stores every conversation turn in a Postgres database so the assistant remembers context across a full session and follow-up questions never feel disjointed
Works across channels without rebuilding, so swapping WhatsApp for Telegram or Slack is a single node change when a client wants a different platform

Tools Used

OpenAI
n8n
WhatsApp

Where to Get Hired for This Skill

On Contra, top freelancers across this stack have earned 138 combined verified reviews from real client projects.

Source: Contra freelancer search · refreshed 26 May 2026

Start Earning as a Freelancer on Contra

Contra is a commission-free professional network for independents. Browse live AI automation work and keep what you earn.

Join Contra Free →

How To Build It

Wire up your three API credentials

Connect OpenAI, the Meta WhatsApp Cloud API using a permanent token, and a Supabase or Postgres database so all three services can communicate through the automation from the start.

Set the client domain as the crawl root

Drop the client's root domain into the workflow variable so the AI agent knows which site to fetch pages from when a customer asks a question, requiring no further configuration per page.

Register the webhook with Meta

Point the Meta WhatsApp Cloud API webhook to the public URL generated by the workflow so every inbound WhatsApp message from the client's customers triggers the automation instantly.

Tune the system prompt for the client's voice

Edit the single system prompt to match the client's brand tone and preferred language, which controls how the bot introduces itself and how formally or casually it replies to every message.

Confirm the bot answers with live data

Send a test message from a real WhatsApp number to verify the bot crawls the client's site, retrieves accurate information, stores the exchange in the database, and returns a coherent reply within seconds.

Pitfalls

Prompt voice drift: if the client rebrands or changes their tone guidelines and you forget to update the system prompt, the bot keeps using the old language indefinitely, which customers notice before you do
Meta 24-hour window violations: if the workflow sends a free-form reply outside the allowed service window instead of switching to an approved template, Meta can restrict the WhatsApp number, taking down the entire client integration
Database connection drops: Supabase free-tier projects pause after periods of inactivity, which causes chat memory writes to fail silently and makes the bot lose conversation context mid-session without throwing an obvious error

FAQ

Can I build this without coding?

Yes. The only configuration is pasting API keys, entering a domain, and editing a text prompt. The crawling, memory, and WhatsApp switching logic are all handled by the workflow without writing a single line of code.

How long does it take?

A clean deployment for a new client takes around 15 minutes once you have the three API credentials ready. Sourcing a Meta WhatsApp Cloud API number and a Supabase database, both available on free tiers, adds roughly 30 to 60 minutes the first time.

What can I charge?

Pricing is entirely your decision and should reflect your market, the client's revenue, and what support tools you are replacing. Many freelancers structure this as a monthly retainer covering hosting, maintenance, and prompt updates rather than a one-time fee.

Which tool is required vs optional?

OpenAI and the Meta WhatsApp Cloud API are required for the core functionality. The Postgres or Supabase memory layer is technically optional but strongly recommended because without it the bot forgets context between every single message, making multi-turn support conversations nearly unusable.

This is original DigiNo analysis. The underlying automation pattern is a community workflow template – view the original on n8n.