How to Test ElevenLabs Agents With TestMu AI

---

ElevenLabs deploys one agent configuration to phone, web voice, and text, with voice conversations running over WebRTC and text-only conversations over WebSocket. TestMu AI connects to the agent on whichever surface it runs on, drives full multi-turn conversations across personas and edge cases, and scores every run.

Before You Begin

---

Before you connect an ElevenLabs agent, make sure you have:

• ElevenLabs account with a published agent
• A number imported under Phone Numbers, for phone agents only
• TestMu AI workspace with agent-testing permissions
• Source docs for scenario generation: prompt, PRD, or knowledge base

Test an ElevenLabs Phone Agent

---

Phone testing covers inbound and outbound callers over native telephony, Twilio, or a SIP trunk. Paste the agent's system prompt to generate scenarios, then add the imported number under Phone Numbers and pick from 100+ voices, background noise conditions, and personas. It covers DTMF entry, transfers, voicemail handling, and carrier latency.

Full setup: Phone agent testing

Test an ElevenLabs Voice Agent

---

Voice testing applies to agents reached over a direct audio stream rather than a phone number. Paste the system prompt to generate scenarios, then create the endpoint profile using WebRTC or WebSocket. It isolates the speech pipeline, so STT and TTS accuracy, turn-taking, and interruption handling get tested without telephony noise in the results.

Full setup: Voice agent testing

Test an ElevenLabs Chat Agent

---

Chat testing applies to agents running in text-only mode, either configured that way or forced with a runtime override. Upload the supporting docs and knowledge base the agent answers from to generate scenarios, then create the endpoint profile using WebSocket. It skips every audio metric and scores logic, tool calls, grounding, and safety, which can run on every commit.

Full setup: Chat agent testing

What You Get With Agent Testing

---

Every ElevenLabs run, on any surface it supports, is scored across:

• 30+ metrics across 8 categories, with configurable thresholds
• Context-aware scenario generation
• Adversarial testing and automated issue detection
• 100+ voices and personas
• Multilingual conversations
• Quality scoring and real-time call monitoring

ElevenLabs-Specific Considerations

---

A few ElevenLabs behaviours are worth building dedicated scenarios around:

• SIP audio runs at G711 8kHz or G722 16kHz regardless of the audio format set on the agent for WebSocket connections. Mismatches here look like quality failures but are codec failures
• Private agents need a signed link for WebSocket or a conversation token for WebRTC. Public agents need only the agent ID
• Text-only mode depends on the agent response event firing, so a silent failure reads as a timeout

Troubleshooting

---

Common failure modes to watch for:

• No transcript returned
• Codec mismatch on SIP calls
• Signed link expired
• Text-only run hangs