Skip to main content

​
The no-code home for your playbook

The Playground collapses a three-day authoring job into about thirty minutes. Describe an agent, get a working playbook, test it by voice, harden it, and take it live - without leaving one screen. It runs the same stack production does. The playbook is authored with SuperDialog; the browser voice call is served by the same AgentRunner that serves a real phone call. What you hear in the Playground is what a caller hears. ▶Open the PlaygroundLaunch it at superdialog.unpod.ai/playground, describe an agent, and start talking to it in your browser. A guest trial lets you build before you sign up.→

​
One screen, three surfaces

1

Agent chat (left, always present)

You describe the agent in plain English (“confirm appointments, fall back to 5pm”). The builder agent authors and edits the playbook for you. Attach a PDF or paste a URL and it reads from those too.
2

Editor tab (right)

The live playbook YAML. Every change the chat makes lands here; you can also edit it by hand. Validation runs as you type.
3

Preview tab (right)

The voice test. Press to talk, speak like a caller, and watch the transcript, the current checkpoint, and the metrics update live.
Preview tab showing the voice pipeline: User speaks into STT, a USER_TURN arrow into DialogMachine.turn() under 'SuperDialog - your flow', an AGENT_TURN arrow into TTS, and 'Caller hears' - with a 'Connected - speak to run a turn through the pipeline' status.

One screen - the builder chat on the right, Editor/Preview tabs up top, the voice profile and model pickers, and the Press-to-talk orb in the center

​
The end-to-end journey

1. Build an agent playbook

Describe it in chat or start from a document; the agent writes the playbook.

2. Test by voice

Press to talk in the browser and confirm it behaves like a real call.

3. Optimize

Run simulated personas, score goal completion, and harden the weak spots.

4. Publish & share

Set visibility, share by link, clone, then deploy - as a voice agent on a number, or as an API endpoint.

​
Reuse an existing agent

You do not have to start from a blank page.

Browse the playbook gallery

The Public tab lets you open and clone any agent others published; the My Playbook tab holds your own drafts.

​
Who this section is for

The Playground is where you build and prove an agent. To put it in front of real callers, deploy the published agent as a Voice Agent on a number (it wires a Speech Pipe for you) - covered in Publish & share.