Note
Access to this page requires authorization. You can try signing in or changing directories.
Access to this page requires authorization. You can try changing directories.
Note
Features in this article are powered by the standard harness, which uses the billing options described in Licensing for agents powered by the standard harness. Learn how to access standard features in Access standard agents and agent flows.
Test your real-time agent directly in Copilot Studio by using voice or text. When you use voice testing, you can speak to your agent and hear its spoken responses. When you use text testing, you can validate the conversation experience for digital messaging scenarios.
Note
Real-time agents are in preview for digital messaging channels. Preview features aren't meant for production use and might have restricted functionality. These features are subject to supplemental terms of use. Microsoft provides these features before an official release so that customers can get early access and provide feedback.
Test modes
| Mode | Channel coverage | Description |
|---|---|---|
| Speech & DTMF | Voice channel | Speak with the agent and hear its responses to validate voice-specific capabilities. |
| Text | Digital Messaging channel | Test the end-to-end messaging experience by sending messages to the agent and reviewing text responses. |
Start a test session
Note
The test panel only supports agents that use the Standard and Premium voice tier (Real-time voice). The test panel doesn't support agents that use other voice tiers for voice testing.
Open the test panel.
Make sure that Real-time is selected as your agent type.
Select the test panel.
In the test panel toolbar, set Chat mode to Speech & DTMF. If Chat mode is currently set to Text, complete the following steps:
- Select the Chat mode dropdown, and then select Speech & DTMF (Preview).
- Save the setting. Your browser might prompt you to allow microphone access. If so, select Allow.
Important
To use voice testing, set your agent's Security > Authentication to No authentication.
Select Start voice conversation.
When connected, you hear the agent's greeting through your speakers or headphones. After the agent finishes speaking, speak your response into the microphone. While the agent is speaking, Talk to interrupt appears above the controls. You can start speaking at any time to interrupt (barge-in). When the agent is listening, the label changes to Listening….
Use the control buttons for in-call actions such as mute or unmute microphone, open DTMF keyboard, or hide the transcript.
To end the session, select End conversation.
To start a new session, select Start voice conversation again.
To switch modes, select the Chat mode dropdown in the test panel toolbar. Switching modes resets the conversation history. You can download the current history by choosing Switch and download snapshot in the confirmation dialog.
Limitations and differences from production
Voice testing in the test panel uses the microphone, speakers, or headset of your device. This is different from the audio hardware and network path used in real phone calls:
- Audio hardware: The test panel uses your device microphone/speakers. Real phone calls use telephone handsets with different sensitivity and noise cancellation.
- Acoustic environment: Your test environment is likely quieter than a real caller's environment.
- Audio path: In the test panel, audio travels directly between your browser and the agent service. Real phone calls pass through the telephony network (PSTN or SIP trunk) introducing latency and compression.
Use caution when relying on test panel results for:
- Barge-in: The barge-in experience depends heavily on the audio hardware and echo cancellation. Barge-in behavior in the test panel might not accurately reflect how it works over a real phone call.
- Voice Activity Detection (VAD) settings: Tuned for telephony audio characteristics.
- Speech quality: Affected by background noise, microphone quality, and audio compression.