Let an agent answer WhatsApp messages to your business number, photos and documents included. Replies go out from the same number, and approved templates reach people who haven't written in the last 24 hours.
- Modes
- Inbound, Outbound
- Set up in
- Dashboard or connectors/*.yaml
- Documentation
- WhatsApp docs
- Publisher
- By connic
Put an agent on your business number
Save a Connection with your Meta app and pick a phone number. When you save an inbound connector, connic sends that number's messages to the agent.
- One Connection per Meta app
- The Connection holds the app ID, app secret, system user token and WhatsApp Business Account ID. Every connector on it picks one phone number of that account.
- Webhook set up on save
- connic subscribes the app to the account and sets the callback URL of the connector's number. Other numbers keep their settings, and deleting the connector removes the callback URL.
- One inbound connector per number
- A number sends its messages to one inbound connector. Outbound connectors can send from the same number.
Get every message verified and ready to read
WhatsApp signs each delivery with your app secret. connic checks it, then starts every linked agent with the text, the sender and any file.
- Verified deliveries
- connic checks X-Hub-Signature-256 against the app secret. A message WhatsApp delivers again maps to the same run.
- Files from the message
- Images, audio, video, documents and stickers arrive in files, with the caption as text. Files over 32 MB are listed in file_issues.
- Sender with or without a number
- from holds the phone number when the person shares it. user_id identifies every sender, and replies reach both.
- Conversation history
- conversation_id identifies one person on one number. Set session.key to input.conversation_id, and follow-up messages share a session.
{
"message_id": "wamid.HBgNNDkxNzYxMjM0NTY3OBUCABIYFDNBRjQ2QkZBMEM2RDQ1NjE3MjAA",
"type": "text",
"text": "Has order ord_8812 shipped yet?",
"from": "+4917612345678",
"user_id": "US.13491208655302741918",
"name": "Jana",
"conversation_id": "whatsapp:9e0f77d9-2fd8-5a8e-8d47-59fbf5b8a091",
"raw": { ... }
}session:
key: input.conversation_id
ttl: 86400Reply within 24 hours, or send a template
Link an outbound connector with the same Connection and number. It sends after the run, when the agent calls it as a tool or when middleware calls it.
- Recipient order
- An explicit to comes first, then the connector's default recipient, then the person who wrote in.
- Text replies
- Up to 4,096 characters, within 24 hours of the person's last message.
- Approved templates
- Outside that window, send template with the name, language and parameters of a template Meta approved for your account.
- Delivery receipts
- The run output holds WhatsApp's message ID. Sent, delivered, read and failed receipts appear in the logs of the inbound connector on the same number.
{
"text": "Yes, ord_8812 left our warehouse today and arrives on Friday."
}{
"to": "+4917612345678",
"template": {
"name": "appointment_reminder",
"language": "en_US",
"parameters": ["Jana", "10:30"]
}
}Declare both directions in code
Declare the inbound and the outbound connector as YAML files in connectors/. Both reference the WhatsApp Connection by name, so the files hold no credentials, and each deploy creates or updates them for its environment.
type: whatsapp
mode: inbound
connection: support-whatsapp
config:
phone_number_id: "106540352242922"
agents:
- whatsapp-supporttype: whatsapp
mode: outbound
connection: support-whatsapp
config:
phone_number_id: "106540352242922"
agents:
- name: whatsapp-support
outbound: { usage: automatic, source: [support-whatsapp] }Where WhatsApp fits
Pick a setup to see what arrives from WhatsApp and what the agent sends back.
A customer sends a photo of an error message, and the agent answers with the fix.
{
"message_id": "wamid.HBgNNDkxNzYxMjM0NTY3OBUCABIYFDVBMjA0N0Q5RTQ2MEI3MUQ4NjAA",
"type": "image",
"text": "The coffee machine shows this error. What should I do?",
"from": "+4917612345678",
"user_id": "US.13491208655302741918",
"conversation_id": "whatsapp:4c2a9e1b-7f3d-5b8e-a6c4-0d2f8e1a3b5c",
"files": [
{
"name": "image-1184303612398871.jpg",
"mime_type": "image/jpeg",
"data": "<base64-bytes>",
"size": 182344
}
]
}{
"message_id": "wamid.HBgNNDkxNzYxMjM0NTY3OBUCABEYEjdCMjA5RjE0QzM2RDU4NEU2AA==",
"wa_id": "4917612345678"
}A customer asks about an order, and the agent looks it up and answers from your number.
{
"message_id": "wamid.HBgNNDkxNzYxMjM0NTY3OBUCABIYFDNBRjQ2QkZBMEM2RDQ1NjE3MjAA",
"type": "text",
"text": "Has order ord_8812 shipped yet?",
"from": "+4917612345678",
"user_id": "US.13491208655302741918",
"conversation_id": "whatsapp:9e0f77d9-2fd8-5a8e-8d47-59fbf5b8a091"
}{
"message_id": "wamid.HBgNNDkxNzYxMjM0NTY3OBUCABEYEjFBM0Q1QzdFOUIyRjRBNkM4AA==",
"wa_id": "4917612345678"
}A daily schedule starts the agent, and it sends an approved template to everyone with an appointment tomorrow.
{
"trigger": "cron",
"schedule": "0 7 * * *",
"triggered_at": "2026-10-14T07:00:00",
"prompt": "Remind every customer who has an appointment tomorrow."
}{
"to": "+4917612345678",
"template": {
"name": "appointment_reminder",
"language": "en_US",
"parameters": ["Jana", "10:30"]
}
}A customer taps Reschedule on the reminder, and the agent offers free slots.
{
"message_id": "wamid.HBgNNDkxNzYxMjM0NTY3OBUCABIYFDhDMkU0QTZCOEQwRjJBNEM2RThBAA==",
"type": "button",
"text": "Reschedule",
"from": "+4917612345678",
"user_id": "US.13491208655302741918",
"conversation_id": "whatsapp:4c2a9e1b-7f3d-5b8e-a6c4-0d2f8e1a3b5c"
}{
"text": "No problem. Thursday 14:00 and Friday 9:30 are free. Which one suits you?"
}WhatsApp FAQ
More connectors
Apache Kafka
Consume topics to trigger agents and produce results back to topics under the source message key.
Inbound · Outbound
Cron Scheduler
Run an agent on any cron expression. connic hosts the scheduler for daily digests, hourly health checks, and weekly audits.
Inbound
GitLab
Run an agent on issues, merge requests and comments in a GitLab project, on gitlab.com or your own instance. It replies with a comment.
Inbound · Outbound
Linear
Run an agent on new and updated issues and comments in a Linear team. It replies in the comment thread or creates a new issue.
Inbound · Outbound
Pipedrive
Run an agent when deals, leads, people or organizations are created, updated or deleted in Pipedrive. It adds its result as a note to the record.
Inbound · Outbound
WebSocket
Stream tokens to the browser over a live socket. Build chat UIs that keep session state across turns.
Sync