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

# Test and publish a voice agent

> Test a voice agent by voice or text with test callers, inputs and tool mocks, then publish a version and manage drafts.

> [!NOTE]
>
> This capability is in **beta**. Behavior can change before general availability.

At the end of this page, you have tested a voice agent and published a version that takes real calls. Testing never edits or publishes the voice agent.

## Before you start

Before you start, sign in to [the dashboard](https://dashboard.hellotars.com) and select the organization you want to work in.

* A voice agent. See [Create a voice agent](https://hellotars.com/docs/guides/voice/build/create-a-voice-agent.md).
* For a voice test, a browser with microphone access.

> [!WARNING]
>
> Test calls use your organization's voice minutes. A voice test reduces the same balance as a real call.

## Test the voice agent

**1. Open the voice agent and select Test**

The test panel opens next to the editor.

**2. Choose Voice call or Text chat**

**Voice call** is a real browser call. **Text chat** checks instructions, transitions, variables and tools without audio.

**3. Choose the version in Test against**

A draft uses the current content after pending edits save. **Live** or another version reproduces that published version.

**4. Optional: choose a test caller**

The **Test caller** dialog sets who this run represents.

**5. Optional: choose where the call starts in Start at**

Leave the default for the normal entry point.

**6. Select Start test call, or type what the caller says**

The live transcript, tool calls and knowledge searches show as the call runs.

[https://dashboard.hellotars.com](https://dashboard.hellotars.com)

![The test panel open beside the Agent page, with Voice call and Text chat, Test against set to the draft, New caller, Inputs, and Start test call](https://hellotars.com/docs/docs/guides/voice/images/voice-test-panel.png)

> [!NOTE]
>
> The panel says tools run normally unless you add a mock. That holds only for read-only tools. A tool that changes data fails in a test call unless it has a mock. See [Test inputs and tool mocks](#test-inputs-and-tool-mocks).

When a voice test ends, select **Open in call history** to see its recording, transcript, tool calls and failure details. Test calls show under **Test** in **Calls**.

### Test callers

| Option | Who the run represents |
| - | - |
| **New caller** | A fresh caller. Fill in **Name**, **Email**, **Phone** and **Company**, or leave them blank for a caller with no profile details. |
| **Saved profile** | A saved profile, copied into a fresh caller. You can save up to 50 profiles. |
| **Returning** | A previous test caller identity, reused. |

### Test inputs and tool mocks

Select **Inputs** to open **Test inputs**. It sets values the test uses in place of a real call. The values affect this run only.

* **Variables** gives variable values for this test.
* **Tool mocks** returns a result you set when a tool runs. Choose the tool, choose whether it **Succeeds**, and give the JSON result.

Tools that change data must be mocked in tests. Read-only tools run for real when no mock matches. To reuse a set, save it with a name and select **Load saved values** later.

### Replay a real call

Open a finished call of this voice agent in **Calls** and select **Open this call in LLM test**. The test panel loads the call's transcript and starting variables. Select **Replay against this agent** to run the same caller turns against the current version, without the microphone.

## Publish a version

> [!WARNING]
>
> Publishing moves every new call on each connected phone number and on the web to the new version. Test the draft in the test panel before you publish.

**1. Select Publish in the editor header**

The **Publish version** dialog opens.

**2. Fix every item under Fix these before publishing**

These issues block publishing. Items under **Worth checking** do not.

**3. Check Where this version goes live**

Each phone number and the web channel show the current version and the new one.

**4. Optional: enter a Version name and describe What changed?**

Both fields are optional.

**5. Select Publish version**

Calls already in progress keep their version.

## Drafts and versions

Select **Drafts and versions** in the editor header to open the panel.

A draft is a separate copy of the voice agent. Editing a draft never touches the live version. Start a new draft from the draft chip in the header, and switch to a draft to test it. **Make live** publishes a draft as the next version.

Every publish creates a version. For each version:

* **Test** runs the version as it is, without rolling back.
* **Rollback** copies the version into what you are editing. The live version stays as it is until you publish.
* **Download this version as JSON** exports it.

## Verify

The **Voice Agents** list shows the voice agent as **Published**. A new call to the voice agent shows the new version in **Calls**.

## Related pages

* [Create a voice agent](https://hellotars.com/docs/guides/voice/build/create-a-voice-agent.md)
* [Distribute a voice agent](https://hellotars.com/docs/guides/voice/launch/distribute-a-voice-agent.md)
* [Review calls](https://hellotars.com/docs/guides/voice/monitor/review-calls.md)
* [Voice billing](https://hellotars.com/docs/guides/billing/voice-billing.md)


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