Test LiveKit agents
- Time to value
- First test report in under 10 minutes
- Setup
- 4 steps
First test report in under 10 minutes
Connect your provider, sync your agents, and validate real calls in one workflow.
- 01
Connect LiveKit
Add your LiveKit API key, secret, and server URL. - 02
Configure rooms
Choose auto-provisioned or customer-controlled rooms. - 03
Run a WebRTC test
Trigger a test room and replay the session after the run.
What you need
- LiveKit API key and secret (LiveKit Cloud > Settings > Keys).
- Room creation plan (auto-provisioned or customer-controlled).
- Optional: enable recording if you want MP4 archives.
Connect in minutes
- 01Go to Agents > Providers > Connect LiveKit.
- 02Enter API key, secret, and server URL.
- 03Select auto or customer-controlled room provisioning.
- 04Run a test room and confirm transcripts populate.
Validation checklist.
Confirm the integration works before you scale your tests.
- Provider shows Connected in Agents > Providers.
- Agents appear in Agents > List with the provider badge.
- A test run produces transcripts and audio in the run summary.
- Test room appears and session replay works after the run.
Built for LiveKit teams.
Provider-aware testing and monitoring, without changing your stack.
WebRTC-only testing
Test LiveKit agents with no phone numbers or SIP setup.Flexible room provisioning
Auto-create rooms or use your own webhook workflow.Replayable sessions
Review LiveKit sessions with full audio and transcripts.
What we measure.
50+ quality metrics across accuracy, conversation quality, voice performance, and task completion.
Accuracy & Correctness
- Factual accuracy
- Intent recognition
- Response relevance
- Hallucination detection
Conversation Quality
- Turn-taking flow
- Interruption handling
- Context retention
- Conversation completion
Voice & Audio
- Latency (time to first word)
- Speech clarity
- Background noise handling
- Accent robustness
Task Completion
- Tool call success rate
- API integration reliability
- Goal completion rate
- Error recovery
Why vendor-neutral
testing?
The same metrics across every provider, instead of scores your vendor reports about itself.
| Aspect | Provider built-in testing | Hamming |
|---|---|---|
| Objectivity | Provider built-in testingOptimized for their platform | HammingVendor-neutral evaluation |
| Consistency | Provider built-in testingMetrics vary by provider | HammingSame 50+ metrics across all providers |
| Cross-vendor comparison | Provider built-in testingCan't compare across vendors | HammingA/B test agents across any provider |
| Independence | Provider built-in testingSelf-reported results | HammingThird-party validation |
| Compliance | Provider built-in testingLimited audit trail | HammingSOC 2 certified, audit-ready reports |
| Scale | Provider built-in testingPlayground-level testing | Hamming50K+ concurrent test calls |
What you get
with Hamming.
- Auto-generate test cases and guardrails from your prompt.
- Pull tool call data, transcripts, and recordings directly from your provider.
- Run your first test in under 10 minutes with 50+ built-in metrics.
- Test both voice and chat agents with unified evaluation.
Frequently asked
questions.
Everything you need to know about testing LiveKit agents with Hamming.
Browse every answer on the FAQ pageConnect your LiveKit API credentials to Hamming, configure room provisioning (auto or customer-controlled), and run automated tests. Hamming simulates real conversations via WebRTC and evaluates agent responses with 50+ quality metrics.
No. Hamming tests LiveKit agents via native WebRTC, eliminating the need for phone numbers or SIP infrastructure. Tests run directly through LiveKit rooms.
Hamming evaluates 50+ metrics including response accuracy, latency, conversation flow, intent recognition, and custom guardrails. All metrics are audio-native with 95-96% agreement with human evaluators.
Yes. Every test run captures full transcripts, audio recordings, and tool call data. You can replay sessions and debug issues with complete conversation context.
Most teams run their first test within 10-15 minutes. Connect your LiveKit API key and secret, configure room settings, and Hamming auto-generates test scenarios from your agent's prompt.
Hamming simulates real user behavior including interruptions, overlapping speech, and mid-sentence changes. Tests validate that agents handle barge-in gracefully without losing conversation context.