Test your Voice Agent
from CI
E2E testing for AI phone agents - real phone calls, real audio, any functional test, via REST API.
$ curl -X POST https://voicegremlin.com/api/runs
-H "Authorization: Bearer vg_xxx"
-d '{"phone_number": "+1-555-123-1234",
"tests": ["Verify that the agent discloses its AI status and audio recording on the first message"]}'
← {"run_id":"abc123","status":"queued"}
$ curl https://voicegremlin.com/api/runs/abc123
← {"status":"complete","passed":1,"tests":[{"passed":true,
"reason": "Agent clearly stated it is an AI"}]}
-H "Authorization: Bearer vg_xxx"
-d '{"phone_number": "+1-555-123-1234",
"tests": ["Verify that the agent discloses its AI status and audio recording on the first message"]}'
← {"run_id":"abc123","status":"queued"}
$ curl https://voicegremlin.com/api/runs/abc123
← {"status":"complete","passed":1,"tests":[{"passed":true,
"reason": "Agent clearly stated it is an AI"}]}
Real phone calls for true end-to-end validation
Building automated tests against a phone agent represents a major side-quest for your DevOps team, but skipping that test is dangerous - if the support line isn't working, how would your customers tell you? VoiceGremlin calls your voice agent's number from a real carrier line so that your e2e test really covers everything: the SIP integration, STT/TTS, and the agent's functional behavior.
LLM as Tester, LLM as Judge
With an intelligent support agent, the user's intent is the test case. This means there's no need for a test case repository or brittle scripted conversation paths - just pass in a plain English description of what you want to validate. Our agent calls yours and role-plays as a customer, and once the call is complete, the transcript is then passed to a Judge LLM which evaluates your agent's behavior against your test goal. The final pass/fail is supplied by inference, not keyword matching or regex.
Ridiculously easy to implement
The VoiceGremlin API was designed around one goal: fast, easy implementation. Your CI or test framework makes two calls: the first one to pass us the phone number and test case, and the second to get the result. There's no SDK and no test case repository to manage. If your tests know how to call an API, the story to build and integrate an e2e test with VoiceGremlin will be a 1-pointer.
Lightweight and un-opinionated
Nothing is more frustrating for an SDET than a test framework that makes you conform to its own idea of how to manage tests. With VoiceGremlin, a test case is just a string that describes what you want to validate and a test suite is just an array of strings. Easy.
Failed tests are actionable
E2E test cases cover a lot of functionality, so when one of them fails it can be frustrating to figure out where to begin troubleshooting. With VoiceGremlin tests, the failure notification includes the complete call transcript; exposing this in your tooling allows the person who sees the failure to begin diagnosing the underlying issue immediately.
Your voice agent is production code. Is it covered by production tests?
100 free minutes. No credit card. First call in under a minute.
Start testing now