WhatsApp Agents
WhatsApp Agents respond automatically to messages from technicians, operators, and supervisors. They consult databases, create work orders, send notifications, and generate reports — all within a WhatsApp conversation.
WhatsApp Agents
When a maintenance technician texts at 2 AM asking about the last vibration reading on compressor C-03, the WhatsApp agent responds in seconds with the exact data — while everyone else is asleep. When an operator reports a leak in pump B-07, the agent creates the work order, assigns it to the on-call technician, and notifies the supervisor, all within the same conversation. That's what a WhatsApp agent does: it brings the intelligence of the platform to wherever the team already communicates.
What is it for?
WhatsApp is where plant personnel actually communicate — it's open, it's fast, and everyone knows how to use it. A WhatsApp agent turns that channel into a 24/7 operational assistant that never sleeps, never loses a request, and never needs to be asked twice.
WhatsApp agents enable:
- Responding to technical queries about equipment status, maintenance history, or part availability at any time
- Receiving failure reports and creating work orders automatically from the conversation
- Sending alerts to specific technicians when the machine monitoring agent detects a critical alarm
- Generating PDF reports on demand during the conversation
- Locating the nearest available technician when there's an emergency at a remote location
How does it work?
The agent receives the incoming message, reads the conversation history for context, and calls the tools it needs to answer: queries a data collection, consults a sensor reading, creates a task, sends a notification. It assembles everything into a clear response in natural language and sends it back to the user.
The agent maintains context: if the technician asked about pump B-12 at the start of the conversation, and then says "when was the last maintenance?", the agent knows the question is about B-12 — there's no need to repeat the equipment name.
How to use it?
Create an agent
- Go to WhatsApp > Agents in the sidebar.
- Click Create Agent.
- Assign a descriptive name — for example: "Maintenance Agent Plant North" or "Equipment Support."
Write the agent prompt
The prompt is the set of instructions that defines how the agent thinks and responds. It has four sections:
| Section | What to write | Example |
|---|---|---|
| Identity | Who the agent is and what it knows | "You are the maintenance assistant for Line A compressors. You have knowledge of equipment C-01, C-02, and C-03." |
| Instructions | How to handle each type of request | "When someone reports a failure, ask for the equipment code, the problem description, and the urgency level before creating the work order." |
| Rules | What the agent must not do | "Never share information about other departments' equipment. Always respond in the same language as the user." |
| Tools | Generated automatically when tools are assigned | — |
Use the Enhance with AI button to have the system optimize each section of the prompt.
Assign tools to the agent
Tools are what allow the agent to act beyond just responding with text:
| Tool type | What it does |
|---|---|
| Query | Searches data collections (equipment specs, inventory, maintenance history) |
| Action | Creates tasks, sends WhatsApp messages, generates reports, locates nearest personnel |
| External | Connects to external APIs, MQTT devices, or PLCs via OPC UA |
Select the tools from the available list and assign them to the agent. For each tool, you can add specific instructions about when and how the agent should use it in this particular context.
Connect a WhatsApp number
Select the WhatsApp number from the Number dropdown. The number must already be registered and connected (see WhatsApp Numbers).
Configure advanced options
| Option | Description | Default |
|---|---|---|
| Welcome message | What the agent says when starting a new conversation | Optional |
| Context messages | How many previous messages the agent remembers as context | 20 |
| Session timeout | Minutes of inactivity before resetting the conversation | 30 minutes |
The context messages setting controls how much conversation history the agent sends to the AI model with each response. A value of 10-15 is sufficient for most use cases and keeps costs lower.
Test the agent
- Send a WhatsApp message to the connected number.
- The agent responds automatically.
- Review the full conversation history in WhatsApp > Conversations — including which tools the agent used and what data it retrieved.
Key benefits
- 24/7 availability without human intervention — the agent works every shift, every night
- Natural language responses — no commands or codes required from the technician
- Full conversation context — the agent remembers what was said earlier in the exchange
- Automatic anti-spam protection — simulates natural typing delays to protect the number
- Work orders and notifications created directly from the conversation without opening the dashboard
- Full conversation log with tool calls visible for supervision and quality improvement
Common use cases
Scenario 1: On-call failure report A technician writes to the agent at 11 PM: "Water leak in pump B-04, area 3, appears serious." The agent confirms the location, asks for a severity estimate, then creates an urgent work order, assigns it to the on-call mechanical maintenance technician, sends them a WhatsApp notification, and responds to the reporting operator: "Work order #MAN-0891 created. On-call technician Carlos Mendez has been notified and will respond within 30 minutes."
Scenario 2: Technical query in the field A new technician is facing compressor C-03 and needs to know the oil change procedure and the recommended lubricant. He texts the agent. The agent searches the equipment's technical specifications collection and responds with the exact lubricant (Mobil Rarus 427, ISO 32), the change interval (every 2,000 operating hours), and the step-by-step procedure. The technician has the information in less than 10 seconds without calling the office or searching through binders.
Scenario 3: Shift report by agent At the end of the shift, the supervisor texts: "Generate the afternoon shift report." The agent queries completed tasks, processed alarms, and inspected equipment for that shift. It generates a PDF with corporate branding and sends it directly in the WhatsApp conversation. The supervisor forwards it to the plant manager. No manual report writing required.