Playground
The Playground lets you make live test calls to yourself (or any number) to verify that a flow sounds and behaves exactly as expected — before using it in a campaign.
Playground calls do not consume your billing quota. Test as much as you need.
Starting a Playground Call
- Go to Flows in the sidebar.
- Find the flow you want to test.
- Click Open Playground (or navigate to the flow and click the Playground tab).
- Enter the phone number to call (your own number or a test number).
- Click Start Call.
Your phone will ring within seconds. When you answer, the AI begins the conversation using your flow's configuration.
During the Call
The playground interface shows the call in real time:
- Transcript — Live text of everything spoken (both you and the AI)
- Call status — Active, ringing, completed
- Duration — How long the call has been running
After the Call
Once the call ends (you hang up or the AI ends the conversation):
| Section | What you see |
|---|---|
| Full Transcript | Complete conversation log with speaker labels |
| Audio Playback | Dual-track recording you can replay |
| Duration | Total call length |
Playground Call History
All your test calls are saved in the playground's call history. You can:
- Browse previous test calls
- Compare transcripts across different flow versions
- Share specific calls with teammates
Super Admins can view the playground history of any user — useful for reviewing how your team is testing flows.
What to Test
Use the playground to validate:
| Scenario | What to check |
|---|---|
| Happy path | Does the AI complete the intended goal of the call? |
| Language | Is the AI speaking in the correct language and accent? |
| Voice quality | Is the voice natural and clearly understandable? |
| Interrupts | If "Allow Interrupts" is enabled, does it handle you speaking mid-sentence? |
| Edge cases | What happens if you say something unexpected or off-topic? |
| Call length | Is the AI ending the call appropriately, or going on too long? |
Iterating on a Flow
After a test call, you might want to adjust the system instruction or initial prompt. The workflow is:
- Note what didn't work in the transcript
- Go to Flows → Edit the flow
- Update the system instruction or initial prompt
- Make another playground call
- Repeat until satisfied
This iteration is free and unlimited.