Testing & chats

The simulator

The simulator lets you run a complete test conversation against an engine before it ever goes live. It speaks for your persona exactly the way the real engine would, so you can read how the conversation flows, check the wording, and confirm the right photos and links land at the right moments.

• Pick a persona first. Use the persona selector in the simulator header — the engine needs an identity to embody, so nothing runs until one is chosen.
• Type as the lead. Send messages the way an incoming contact would, and the persona replies turn by turn.
• It tests the engine you're building. The simulator runs against the engine on your canvas, so you can tweak the flow and re-test without publishing anything.

Outreach-context mode

Outreach-context mode has the AI compose the first proactive message — the cold opener — instead of waiting for the lead to speak. This mirrors the live OUTREACH action your bot triggers in production.

• Attach a photo or screenshot. The AI actually “sees” the image — it reads what’s in it (a profile, a story, a post) and writes an opener that references it.
• Add optional text. You can include a short note for extra ignition context, but the screenshot alone is enough.
• The result is the same opener the engine would send when starting a brand-new chat from scratch.

Follow-up

The Follow-up button triggers a re-engagement message on an existing test conversation — the nudge the engine would send when a lead goes quiet. It runs the same logic as the live follow-up action, so the per-conversation cooldown and cap apply here too. If a follow-up is on cooldown or has hit its limit, no message is sent and the simulator tells you to wait.

Event-trigger injection

Use Event Trigger to simulate a platform event — for example a story_reaction — so you can test engine Event Trigger start points without waiting for one to happen for real. Pick the event, add an optional bit of context (like what the reaction was), and send it to see how the engine kicks off from that signal.

Chat history & transcripts

Every conversation — simulated or live — is recorded as a read-only transcript you can reopen any time from your chats. It shows each message with timestamps and any photos that were sent, laid out persona-vs-lead.

• Timestamps and photos. Messages carry their send time; photos render inline, with a clear “photo unavailable” indicator if an image can’t be loaded.
• The trace uses your own labels. Alongside a chat you can see the engine’s reasoning trace — but only your engine’s phase and card labels, the facts it extracted, and what it understood from any image. Our internal plumbing is hidden by design.

Sessions & leads

A session is one conversation between a persona and a single lead. In the simulator the lead is always simulator-user, so you can recognise your test runs at a glance. Live sessions are tied to whoever your bot is talking to.

Report a problem — copy the conversation ID

There’s no in-app bug form. To report an issue with a specific conversation, copy its Conversation ID — the copy control sits in the simulator header and in the transcript header — and send it to support. Support uses that ID to look up the full diagnostics for that conversation server-side, so they see exactly what happened without you having to paste anything sensitive.

Reset

Reset clears the current test conversation from the simulator so you can start fresh. It is non-destructive — it does not delete the stored session, so the transcript is still on record if you need it.

Next steps

Once a conversation reads the way you want it to, you’re ready to connect it. See building engines to refine the flow, going live to wire it to your tools, and tokens & billing for how usage is metered.