OpenClaw + WhatsApp + CRM: Build a Lead Capture System That Never Sleeps

Digital marketing agency (UK): Three agents, all needing WhatsApp access — one for client intake, one for campaign reporting, one for billing inquiries. Each needs its own number or sub-account.

The demand is real. The technical barrier is knowing exactly how to wire this together.

WhatsApp Connection Options and Tradeoffs

WhatsApp is not like Slack or Telegram. You cannot simply create a bot account and connect it to an API. There are three practical ways to connect OpenClaw to WhatsApp, each with meaningful tradeoffs.

Option 1: WhatsApp Business API

The official path. Meta requires business verification (usually 1–3 business days) and you send messages through a BSP (Business Solution Provider) or directly through the Cloud API.

In OpenClaw, this connects via the webhook channel type. Meta sends incoming messages to a webhook URL on your OpenClaw instance; your agent sends outbound messages through the Meta Graph API.

# OpenClaw environment config for WhatsApp Business API
WHATSAPP_PHONE_NUMBER_ID=your-phone-number-id
WHATSAPP_ACCESS_TOKEN=your-permanent-access-token
WHATSAPP_VERIFY_TOKEN=your-webhook-verify-token
WHATSAPP_WEBHOOK_PATH=/webhook/whatsapp

Messaging limits start at 1,000 unique users per day and can scale to unlimited with a verified business account and a good sender reputation.

Option 2: Twilio WhatsApp Sandbox / Production

Twilio provides a WhatsApp sandbox for development and a production route once your Twilio account is approved. This is a good option for teams already in the Twilio ecosystem.

# OpenClaw Twilio WhatsApp config
TWILIO_ACCOUNT_SID=your-account-sid
TWILIO_AUTH_TOKEN=your-auth-token
TWILIO_WHATSAPP_NUMBER=whatsapp:+14155238886

Cost runs $0.005 per message sent plus Twilio's per-message fee. Predictable and well-documented.

Option 3: WhatsApp Bridge (Baileys)

The OpenClaw community has built bridges using the Baileys library, which implements the WhatsApp Web protocol. This lets you connect OpenClaw to a regular WhatsApp account by scanning a QR code — no business verification required.

This is useful for internal use cases (the NBFC group report scenario, for example) where you're reading messages from your own WhatsApp number, not operating as a business bot. Be aware that this method violates Meta's Terms of Service for commercial bot use.

The Lead Capture Architecture

Here is the complete data flow for a WhatsApp lead capture system:

Inbound WhatsApp message
        │
        ▼
OpenClaw agent (qualification loop)
        │
        ├── Asks qualifying questions (2–4 exchanges)
        │
        ├── Lead scored / categorized
        │
        ├── [Tool call] → CRM write (HubSpot / Close.io / Notion)
        │
        ├── [Tool call] → Gmail: send confirmation to lead
        │
        └── [Tool call] → Gmail: alert sales rep if hot lead

The key insight: OpenClaw is not just a chatbot here. It is executing a multi-step workflow where each tool call is an API request to an external system. The agent decides when to make each call based on where it is in the qualification flow.

A properly designed openclaw crm automation agent does four things:

1. Collects structured data — name, company, requirement, budget, timeline
2. Scores the lead — based on criteria you define in the system prompt
3. Writes to CRM — via the CRM skill or a custom webhook skill
4. Notifies the right people — immediate email/Slack alert for hot leads, daily digest for cold ones

Setting Up the OpenClaw Agent: Channel Config and System Prompt

Once your WhatsApp channel is connected, you configure the agent with a system prompt that defines its qualification behavior.

Here is a production-ready system prompt for a facilities management company:

You are a professional assistant for [Company Name], a facilities management company serving commercial properties in [Region].

Your job is to qualify incoming service requests from property managers and building owners.

QUALIFICATION FLOW:
1. Greet the contact warmly and confirm they are reaching [Company Name].
2. Ask for their name and the property they are inquiring about.
3. Ask what type of service they need. Common categories: HVAC, electrical, plumbing, general maintenance, cleaning, security systems.
4. Ask the approximate size of the property in square feet or number of floors.
5. Ask whether this is an urgent issue (same-day response needed) or a scheduled request.
6. Ask for their preferred contact method for the quote (WhatsApp or email).

QUALIFICATION CRITERIA:
- Hot lead: Commercial property 10,000+ sqft OR urgent request → notify team immediately
- Warm lead: Residential or <10,000 sqft, non-urgent → log in CRM, follow up within 24h
- Out of scope: Requests outside [Region] or for services you don't offer → politely explain and redirect

AFTER QUALIFICATION:
- Summarize what you've collected: "Just to confirm, you need [service] for [property] — does that sound right?"
- Tell them: "Our team will send you a detailed quote within [timeframe]. You'll receive it on WhatsApp or email, whichever you prefer."
- Do NOT provide pricing. Do NOT make commitments about availability.

TOOLS AVAILABLE:
- create_crm_contact: Use after qualification is complete
- send_email_alert: Use for hot leads only
- send_whatsapp_confirmation: Use to send the lead a confirmation message

TONE: Professional, concise, warm. Response length: 1–3 sentences per message. Do not use bullet points in WhatsApp messages.

This prompt does several things that matter in practice:

• It constrains the agent to a defined flow (not an open-ended chat)
• It specifies when to use each tool call (after qualification, not during)
• It gives explicit scoring criteria so the agent classifies leads consistently
• It handles out-of-scope requests gracefully rather than hallucinating capabilities

Example Conversation Flow: Qualification to CRM Handoff

Here is what the actual conversation looks like from the lead's perspective:

Lead:    Hi, I need maintenance services for my building

Agent:   Hi! Thanks for reaching out to BuildRight Facilities.
         I can help get you connected with our team.
         Could I get your name and the name or address of the property?

Lead:    This is Sarah Chen, property at 45 Marina Bay Tower, commercial building

Agent:   Thanks, Sarah. What type of service do you need for Marina Bay Tower?
         (For example: HVAC, electrical, plumbing, cleaning, or general maintenance)

Lead:    Our HVAC system has been making a loud noise since this morning
         and one floor is not getting cold air

Agent:   Got it — that sounds like it could be an urgent issue.
         How large is the property approximately, in floors or square footage?

Lead:    It's 18 floors, about 120,000 sq ft

Agent:   And would you prefer to receive the quote on WhatsApp or by email?

Lead:    WhatsApp is fine

Agent:   Perfect. Just to confirm: you need urgent HVAC service for Marina Bay Tower
         (18 floors, ~120,000 sqft) — does that sound right?

Lead:    Yes exactly

Agent:   Our team has been notified and will contact you within 2 hours.
         You'll receive a quote directly on WhatsApp. Thank you, Sarah.

[BACKGROUND: Agent calls create_crm_contact with all collected data,
 then calls send_email_alert to the on-call HVAC team lead]

The lead experience is clean and professional. The back-end work — CRM entry, team alert, confirmation message — happens silently in the same turn.

CRM Write-Back: Pushing Qualified Leads to HubSpot, Close.io, or Notion

OpenClaw's skill system lets you define custom tools that make API calls. The openclaw crm automation integration is a skill configured with your CRM's API credentials.

HubSpot Integration

The HubSpot skill calls the Contacts API to create or update a contact:

{
  "skill": "crm_hubspot",
  "config": {
    "api_key": "your-hubspot-private-app-token",
    "default_pipeline": "Sales Pipeline",
    "default_stage": "New Lead",
    "lead_source": "WhatsApp"
  }
}

When the agent calls create_crm_contact, it passes the structured data it collected:

{
  "firstname": "Sarah",
  "lastname": "Chen",
  "phone": "+65XXXXXXXX",
  "company": "Marina Bay Tower",
  "property_size": "120000",
  "service_type": "HVAC - Urgent",
  "lead_score": "hot",
  "source": "WhatsApp"
}

HubSpot creates the contact, assigns it to the pipeline, and any automation rules you've configured in HubSpot (task assignment, sequence enrollment) fire automatically.

Close.io Integration

Close.io is common in sales-heavy businesses. The skill hits the Leads endpoint:

{
  "skill": "crm_close",
  "config": {
    "api_key": "your-close-api-key",
    "status_id": "stat_new",
    "lead_source": "WhatsApp Inbound"
  }
}

Notion as a Lightweight CRM

For smaller operations that use Notion as their source of truth, the Notion database integration works well:

{
  "skill": "notion_database",
  "config": {
    "database_id": "your-notion-database-id",
    "api_token": "your-notion-integration-token"
  }
}

The agent creates a new page in your Leads database with properties mapped to each qualification field. Works with any Notion database schema you've already built.

Follow-Up Sequence Automation

A one-time qualification is useful. A follow-up sequence is where deals are actually won.

After the initial CRM write, you can configure a follow-up chain:

Day 0: Quote sent via WhatsApp (confirmation message)
Day 1: Check-in if no response — "Did you receive our quote?"
Day 3: Value-add message — "We've done HVAC work at [similar building]. Happy to share a reference."
Day 7: Final follow-up — "Our availability for next week is limited. Want to schedule a site visit?"

In OpenClaw, this is handled with the cron automation feature. A scheduled job checks for leads in the "Awaiting Response" stage and triggers the appropriate follow-up message based on how many days have elapsed since contact.

Quote and Proposal Generation for Service Businesses

The facilities management use case includes quote generation — a step beyond simple lead capture.

This requires the agent to have access to your pricing data. There are two approaches:

Approach 1: Knowledge base (simpler) Load your pricing sheet into the agent's context as a structured knowledge file. The agent references it to calculate quotes based on property size and service type. Works well for standardized pricing.

Approach 2: External calculation API (more accurate) Build a small API endpoint that accepts the qualification parameters and returns a calculated price. The agent calls this as a tool. Useful when pricing has complex variables (regional rates, material costs, labor tiers).

The agent then generates a quote in a structured format and sends it via WhatsApp using the WhatsApp message template system (for Business API users) or as a formatted text message.

Hi Sarah, here is your HVAC service quote for Marina Bay Tower:

Service: Emergency HVAC diagnosis + repair
Property: 18 floors / ~120,000 sqft
Response time: Within 2 hours

Estimated quote: SGD 2,800–3,400
(Final price confirmed after site assessment)

This quote is valid for 7 days.
To confirm, reply YES and we'll schedule immediately.

— BuildRight Facilities Team

When Sarah replies "YES", the agent captures the confirmation, updates the CRM deal stage to "Closed/Won - Pending Confirmation", and sends an internal alert to schedule the site visit.

Monitoring and Uptime: The Most Overlooked Requirement

WhatsApp sessions have a specific failure mode that catches teams off guard.

For unofficial bridge methods (Baileys/WhatsApp Web), the QR code authentication session expires. Sometimes after a few days, sometimes after a few hours if there's a connectivity interruption. When it expires, the agent stops receiving messages — silently. No error. No alert. Just dropped leads.

For the official Business API, the webhook can fail if your server goes offline, returns errors, or if your Meta access token expires. Meta will attempt delivery retries for a limited window, then stop trying.

The monitoring requirements for a production WhatsApp agent:

1. Session health check — automated check every 5–15 minutes confirming the WhatsApp connection is active
2. Heartbeat monitoring — if the agent hasn't processed a message in X hours during business hours, alert
3. Automatic restart — when the agent crashes or the session drops, it should restart and re-authenticate without manual intervention
4. Backup notification — if WhatsApp drops, critical inbound messages should trigger a fallback (SMS, email)

This is precisely what makes managed hosting the right choice for production WhatsApp agents. A self-hosted OpenClaw instance requires you to configure all of this yourself — uptime monitoring, alerting, automatic restart on crash, token refresh before expiry.

On Clawfleet, session health monitoring is built into every instance. You get alerted if your WhatsApp connection drops, and the platform handles restarts automatically. The uptime SLA covers the infrastructure — your job is to write the agent prompt.

Putting It Together: The Full Stack

To summarize the complete openclaw whatsapp integration + openclaw crm automation stack:

Infrastructure layer:

• OpenClaw instance (hosted on Clawfleet)
• WhatsApp connection (Business API recommended for production)
• Domain with HTTPS for webhook reception

Agent layer:

• System prompt defining qualification flow and scoring criteria
• Tool definitions for CRM write, email alert, WhatsApp confirmation
• Follow-up cron job for sequence automation

Integration layer:

• CRM: HubSpot, Close.io, or Notion (via skill config)
• Email: Gmail skill (for agent-sent confirmations and alerts)
• Optional: Slack skill for real-time sales team alerts

Monitoring layer:

• Session health checks (built into Clawfleet)
• Lead volume alerting (zero leads in X hours during business hours = something is wrong)
• Daily digest of leads captured and their status

The NBFC in Mumbai deployed this stack to read WhatsApp group reports from their collections team. Their agent parses the structured data in each message, looks up the corresponding loan record via API, flags discrepancies automatically, and posts a daily reconciliation summary. What previously took a junior analyst 3 hours each morning now takes 4 minutes.

The facilities management company in Dubai processes 40–60 inbound WhatsApp inquiries per week. Their agent qualifies and logs every one of them. Their sales team arrives each morning to a CRM dashboard with scored leads, already segmented into hot/warm/cold, with the qualification notes already attached to each contact.

Neither company is technical. Neither wrote code to build this. They configured an agent, connected the integrations, and deployed it.

That is the actual value proposition of OpenClaw for business owners — not the AI, but the automation layer that connects the AI to the systems you already use. WhatsApp is where your customers start. Your CRM is where your business runs. OpenClaw is the agent that bridges them.