Inbound WhatsApp messages follow the same durable text session path as all other text channels, with conversations keyed on the customer and agent phone pair.
The platform supports inbound WhatsApp as a text channel. WhatsApp messages are processed through the same durable text session path used by web, email, SMS, and iMessage conversations. This means WhatsApp conversations create entries in the world model, appear in the standard conversation list, and benefit from the same session continuity, memory, and context graph capabilities as every other channel.
How It Works
When a customer sends a WhatsApp message to an agent's business phone number, the platform:
Resolves or creates a conversation keyed on the (customer phone, agent phone) pair. Repeat messages from the same customer to the same agent phone resume the existing conversation rather than starting a new one - matching the behavior of SMS and iMessage.
Drives a turn through the durable text session, including context graph traversal, reasoning, memory access, and any configured skills or integrations.
Returns the agent's reply synchronously to the messaging service, which delivers it back to the customer over WhatsApp.
Because WhatsApp uses the same session path as other text channels, all platform capabilities work identically - approval-gated integration writes, escalation, surfaces, and conversation history are all available.
Response Modality
Each WhatsApp number can be configured with a response modality - either voice note or text - controlling how the agent replies on that number.
When a number is configured for text responses:
Inbound text messages are processed through the text turn pipeline, and the agent replies with text.
Inbound voice notes on the same number receive a brief reply directing the user to send text instead.
When a number is configured for voice note responses:
Inbound voice notes are processed through the voice turn pipeline, and the agent replies with audio.
Inbound text messages on the same number receive a brief reply directing the user to send a voice note instead.
Text turns use the same session continuity and agent pipeline as voice turns. Conversation state persists across turns for the same speaker, so multi-turn text conversations maintain full context. Concurrent messages from the same speaker are serialized - if a second message arrives while the first is still being processed, it is rejected to prevent race conditions in the conversation state.
Voice Notes
When a patient sends a voice note on a number configured for voice note responses, the platform transcribes it, runs the full reasoning engine pipeline, synthesizes the agent's reply as spoken audio, and returns it as a voice note in the same thread. Patients hear a natural spoken response without making a phone call.
Switching between text and voice notes in the same thread preserves full conversation context. Concurrent voice notes from the same patient are serialized to prevent race conditions. Phone numbers are E.164-normalized so international numbers work correctly regardless of how the messaging provider formats them.
Conversation Continuity
Each WhatsApp thread is identified by the combination of the customer's phone number and the agent's business phone number. This composite key ensures that:
A returning customer picks up where they left off without repeating context.
Different agent phone numbers maintain separate conversation threads, even for the same customer.
Conversation history is visible in the standard conversation list and detail views alongside web, email, SMS, and iMessage conversations.
The thread persists until it is explicitly closed - there is no reset keyword and no automatic expiry. To restart a customer on a clean thread, close the conversation through the conversations API; see Starting a Fresh Conversation.
Channel Behavior
WhatsApp conversations behave like other text channels:
No acoustic emotion data. Emotion detection models that rely on audio (prosody, vocal bursts) do not apply. Language-based emotion detection remains active.
Approval gates enforced. Integration writes configured with approval policies are gated on WhatsApp turns, just as they are on SMS and web.
Long-running tools reply once. When a reply depends on a background tool, the intermediate acknowledgement is suppressed and the real answer is delivered as a single follow-up message once the tool finishes - the same asynchronous-channel behavior as email, SMS, and iMessage. See Long-Running Tools and Reply Delivery.
Turn persistence. If turn persistence fails during a WhatsApp conversation, the platform surfaces the failure explicitly rather than returning a success response with silently lost data.
Last updated
Was this helpful?