> ## Documentation Index
> Fetch the complete documentation index at: https://docs.botcadence.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Playground

> Test Chat and Voice workflows before deploying them to a live channel

**Build → Playground** is a safe test area for your project. Choose a chat or voice workflow, try a conversation, and fix problems before a website visitor, caller, or campaign recipient can reach your AI Agent.

## What you can test

* **Chat workflows** — send messages, inspect the agent's replies, and verify knowledge retrieval, variables, tools, and handover behavior.
* **Voice workflows** — start a voice test, speak naturally, and check turn-taking, interruptions, audio settings, and the path to completion.
* **Customer context** — when the test surface asks for a customer profile, use a known profile to validate personalization and the unified customer record.
* **Workflow selection** — the workflow picker distinguishes **Chat** and **Voice** flows. If a flow is missing, confirm its type and that it belongs to the current project.

## Run a test

**1. Open Playground**

From the project sidebar, select **Build → Playground**. The page loads the workflows available to the project.

**2. Choose a workflow**

Select the Chat or Voice workflow you want to exercise. Use the edit action beside the selected workflow if you need to adjust its nodes.

**3. Test the happy path**

Start with the normal customer journey: greeting, intent discovery, knowledge answer, and the expected completion action.

**4. Test failure paths**

Try an unknown question, missing required information, a tool error, a request outside scope, and an explicit request for a human.

## A useful test matrix

| Scenario | What to verify |
| - | - |
| Known question | The response uses the right source and does not overstate certainty. |
| Unknown question | The agent says what it cannot verify and follows the fallback or handover path. |
| Tool action | Required inputs are collected, writes are confirmed, and failures are explained. |
| Handover | The conversation reaches the intended human queue with useful context. |
| Long or interrupted turn | The agent keeps the conversation state and does not duplicate an action. |

## Before going live

1. Confirm the workflow is the one attached to the intended [channel](https://docs.botcadence.com/docs/deploy/channels.md), [inbound](https://docs.botcadence.com/docs/deploy/inbounds.md), or [campaign](https://docs.botcadence.com/docs/outreach/campaigns.md).
2. Verify the required [knowledge sources](https://docs.botcadence.com/docs/build/knowledge-base.md), integrations, and workflow tools are configured.
3. For voice, preview the selected TTS setup in [Voice Lab](https://docs.botcadence.com/docs/build/voice-lab.md).
4. Publish the workflow from the designer. Playground can exercise a draft, but live traffic uses the published version.
5. Send a real, low-risk test through the target channel and confirm the resulting thread in [Conversations](https://docs.botcadence.com/docs/operate/conversations.md).

> [!WARNING]
>
> Do not use real customer secrets or irreversible production actions while testing. Use test records and destinations, especially for email, messaging, order updates, and CRM writes.

## If a workflow is not listed

Check that you are in the right organization and project, that the workflow is a Chat or Voice workflow rather than a template, and that the page has finished loading. Create a workflow under [Workflows](https://docs.botcadence.com/docs/build/workflows.md) if the project has none.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.