VoiceGremlin gives teams a practical way to run end-to-end checks against AI phone agents from the same pipelines that test the rest of their product. It places real carrier calls, follows plain-language test goals, and returns evidence that is useful when a test passes or fails.

What's included in this release:

• Run tests through real phone calls so SIP, speech recognition, speech synthesis, and agent behavior are checked together.
• Describe each expected behavior in plain language instead of maintaining brittle scripted conversation paths.
• Evaluate completed conversations with an LLM-based judge that returns a clear pass or fail for each goal.
• Review full call transcripts with failure reasons so troubleshooting can start from the exact conversation.
• Connect CI systems through a simple REST workflow that queues a batch and retrieves the completed results.

Together, these capabilities turn voice-agent testing into a repeatable pipeline step while keeping test definitions lightweight and the output actionable.