You build a WhatsApp chatbot in Stravax Engage by dragging nodes onto a visual canvas (send a message, ask a question, branch on a condition, and more), connecting them with edges, and publishing. Once published, a matching inbound message starts a flow session on that conversation, and the flow steps forward one node at a time as the contact replies.
Open the flow builder and drag nodes from the palette onto the canvas, then connect them with edges to define the order the conversation follows. Configure each node in the side panel (message text, question, condition), then publish. A draft never triggers, no matter how complete it looks; only a published flow runs.
A flow starts when an inbound WhatsApp message matches its configured trigger:
| Trigger | Starts the flow when |
|---|---|
| Keyword | The inbound message matches a keyword you set |
| Click-to-WhatsApp | The contact arrived from a click-to-WhatsApp ad |
| New conversation | Any first message on a brand-new conversation |
If more than one trigger could match the same message, keyword takes priority, then click-to-WhatsApp, then new conversation.
At a feature level, the palette covers:
A safeguard stops a looping flow from running forever: if a path cycles without reaching an end, the session closes automatically after a capped number of steps.
While a flow session is active on a conversation, the flow owns the reply: away/welcome auto-replies and the AI qualifier both stand down for that conversation. If a human agent sends a manual reply, that ends the flow session immediately, handing the conversation back to normal human control.
Yes. Editing and republishing doesn't disrupt conversations in progress: a session that started before the update keeps running on the version it started with, and only new sessions pick up your changes. Each plan caps how many flows can be live at once; see current plans for details and upgrade options.
Use the test-on-WhatsApp option to message from your own number and walk through the published flow live. If you test against a conversation that already has a live flow session, the test is refused rather than ending that session, since it would interrupt a real conversation. Use a separate WhatsApp number to avoid the conflict.
The funnel view shows how many sessions started and where people dropped off, step by step, with a badge on each node showing how many stalled there. The logs panel shows what happened at each step for a given session, useful for tracing why a contact ended up where they did.