Automate WhatsApp lead capture and replies with Whapi, Ollama and Sheets
Receive WhatsApp messages via Whapi, generate AI replies with a local Ollama model, log conversations in Google Sheets, and auto-capture leads — all without touching a cloud LLM.
This n8n template builds a fully automated WhatsApp AI CRM using Whapi.cloud for messaging and Ollama for 100% local AI inference — no OpenAI costs, no data leaving your server.
# How it works
A Webhook node receives inbound WhatsApp messages from Whapi.cloud.
A Code node extracts the sender's phone, name, message text, and filters out outbound/non-text messages.
An IF node ensures only real inbound text messages from customers are processed.
Google Sheets is used to fetch that customer's full conversation history, enabling memory across sessions.
A Code node builds a full prompt — system instructions + conversation history + new message — passed to the AI model.
Ollama (via LangChain LLM Chain node) generates a contextual reply using a local model (default: gemma3:1b).
The user message and AI reply are each appended to Google Sheets as conversation history logs.
A separate Google Sheets upsert captures or updates the lead record with phone and name.
The AI reply is sent back to the customer via Whapi's HTTP API.
# How to use
Set up a Whapi.cloud account and connect a WhatsApp number. Point the webhook to your n8n webhook URL.
Create a Google Sheet with a History tab (columns: Phone, Name, Role, Message, Timestamp) and a Leads tab (columns: Phone, Name, CreatedAt).
Add your Google Sheets credentials and replace YOUR_GOOGLE_SHEET_ID in the relevant nodes.
Run Ollama locally or on your server. Pull the model: ollama pull gemma3:1b. Update the model name in the Ollama node if using a different model.
Customise the system prompt inside the Build AI Prompt node to match your business (real estate, support, bookings, etc.).
Activate the workflow and send a WhatsApp message to test.
# Requirements
Whapi.cloud account (WhatsApp Business API)
Ollama running locally or on a self-hosted server
Google Sheets (with OAuth2 credentials connected in n8n)
# Customising this workflow
Switch AI models: Swap gemma3:1b for any Ollama-supported model like llama3, mistral, or phi3 depending on your hardware.
Change the industry: Edit the system prompt in Build AI Prompt to serve any business — bookings, customer support, sales qualification, etc.
Upgrade the CRM: Replace Google Sheets with Airtable, Notion, or a real CRM (HubSpot, Pipedrive) by swapping out the Sheets nodes.
Add handoff logic: Insert a condition to escalate to a human agent if the message contains keywords like "speak to someone" or "human".
Multi-language: The system prompt already instructs the AI to reply in the customer's language — no extra setup needed.
# Who is this for
It's designed for service businesses (real estate, consultants, agencies) that want to respond to inbound WhatsApp leads instantly, log conversations, and build a simple CRM — all from a single workflow.
Early Deal