Test your agent
The Playground is where you talk to your agent before it talks to a single real contact. You play the prospect, the agent replies with the same prompt, the same knowledge base and the same engine as in production, and nothing leaves WhatSetter: no WhatsApp message, no lead, no meeting.
Open the Playground
Section titled “Open the Playground”- Open Playground in the sidebar.
- Under Choose a campaign to test, pick a campaign. Active campaigns come first, the others under Not active.
- Pick what you want to test (next section) and start writing.
From the prompt editor, the Test button saves your draft and opens the Playground.
Choose what to test
Section titled “Choose what to test”At the top of the page, What do you want to test? offers up to four cards:
| Card | What it tests |
|---|---|
| AI Behavior | The lead writes first. You test how your agent replies, qualifies and handles objections. |
| First Message | Your agent writes first. You test the opener it sends to a new contact. |
| Follow-up, Cold Lead | The reminder sent to a lead who never replied. |
| Follow-up, Abandoned | The reminder sent after a conversation went silent. |
The follow-up cards only appear when the campaign has a follow-up prompt. If the campaign runs several A/B variants, the card shows how many, and you can pin the variant you want to test.
The left pane shows the exact prompt being tested, with your campaign variables already filled in. Edit in the editor takes you to the prompt editor.
Simulate a conversation
Section titled “Simulate a conversation”Write in the field at the bottom as if you were the lead, or tap one of the example messages. Your message gets the WhatsApp ticks, the agent shows as typing, then replies in one or several bubbles, exactly as it splits messages in production.
Above the chat, a pill tracks what the agent decided about this lead: Pending, Qualified, Not qualified or Booked. Qualification starts after a few exchanges, so do not expect a verdict on the first reply.
Test the first message
Section titled “Test the first message”- Pick the First Message card.
- Click Generate First Message.
- Enter a Lead First Name and confirm.
The message is generated from your first-message template, with {{first_name}} replaced by the name you typed. Variables from your lists ({{lead:column}}) are filled with a sample contact taken from your latest lists; a variable no contact has is shown as a visible marker, the same way it would be on a real send. See Campaigns for the variables.
Test the follow-ups
Section titled “Test the follow-ups”Follow-ups are triggered by hand: there is no clock to fast-forward.
- Pick Follow-up, Cold Lead or Follow-up, Abandoned.
- Chat with the agent. For an abandoned conversation, the lead has to reply at least once first (Reply as the prospect first).
- In the Follow-ups bar under the chat, choose Cold Lead or Abandoned, the step, and click Generate.
A divider (Lead never replied or Conversation went silent) marks where time is supposed to have passed, and the follow-up appears as an amber bubble. If follow-ups are off for this campaign, the bar says so and links to the editor. See Automatic follow-ups.
Test the booking without booking
Section titled “Test the booking without booking”When the campaign has in-chat booking, a Booking switch appears above the chat, with the note simulation · no booking created.
- The agent reads your real availability and proposes real slots.
- When the lead picks one and gives what the calendar needs, the agent confirms and the pill switches to Booked.
- No meeting is created in your calendar, ever.
Switch Booking off to see how the agent behaves when it only Sends the link.
What is real and what is not
Section titled “What is real and what is not”| Real, same as production | Not real |
|---|---|
| The prompt, the variant, your style | No WhatsApp message is sent |
| The knowledge base | No lead is created in your CRM |
| Qualification and status decisions | A/B statistics are not touched |
| Message splitting into bubbles | No meeting is created in your calendar |
| Real calendar availability | Sending delays are shortened |
Restart, history and export
Section titled “Restart, history and export”- Restart clears the chat and starts a new test. The previous test stays in the history.
- Test history lists the last 20 tests of the campaign, saved automatically after each reply. Open one to continue it.
- Export downloads the conversation as a JSON file.
AI credits
Section titled “AI credits”Every reply, first message or follow-up generated here uses AI credits from your monthly allowance. When they are used up, the field shows No credits remaining and testing stops until the next month or a plan change. See Plan & billing.