VoiceGremlin
VoiceGremlin is a SaaS tool for automated end-to-end testing of intelligent voice agents that answer phone calls. It places real carrier calls to a customer's voice agent, role-plays a customer using an LLM, and uses an LLM judge to return pass/fail results and transcripts for CI pipelines.
At a Glance
- Software engineering and SDET teams
- Companies deploying intelligent voice response/virtual assistant systems
- Teams integrating customer-support, airline, hotel, restaurant, or other phone agents into CI/CD
- Developers who need vendor-agnostic testing rather than a voice-agent hosting platform
AI Tools by VoiceGremlin
(1)
VoiceGremlin
Voice Agent Testing via REST API
Discussions
No discussions yet
Be the first to start a discussion about VoiceGremlin
Latest News
Show HN launch: VoiceGremlin announced as a working SaaS tool for automated tests against AI phone agents, with a free trial and LLM-as-judge pass/fail pipeline.
Founder launch write-up explained the product rationale, approximately 45-day build process, design goal of doing one thing well, and underlying stack.
Public CI repository added a Python CI script and updated/renamed the Bash and PowerShell examples.
Initial public commit published for the VoiceGremlin CI scripts repository under an MIT license.
Products & Services
Usage-based REST API for queuing batches of natural-language tests against a phone number and polling run results. It calls the target voice agent, supplies a synthetic caller, evaluates the transcript with a judge LLM, and returns Boolean pass/fail results, reasons, and transcripts.
Open-source MIT-licensed Bash, PowerShell, and Python scripts for adding VoiceGremlin checks to CI/CD pipelines; they poll until completion and exit with pass/fail/error codes.
Market Position
VoiceGremlin positions itself as a focused, lightweight, vendor-agnostic API for true phone-line end-to-end testing of voice agents—essentially a 'Mailosaur for phone agents.' It differentiates from building an internal harness by packaging telephony, audio streaming, the calling LLM, and judge LLM behind two API calls, while deliberately avoiding test-case management, voice-agent hosting, SDKs, and upsell features.
Leadership
Founders
Nick Hall
Founder and solo developer; a former SDET/software-engineering middle-manager with a background in test automation. He describes the product as informed by experience using Mailosaur and by the difficulty of automating phone-agent tests.
Executive Team
Nick Hall
Founder and solo developer
Former SDET and software-engineering middle-manager with a background in test automation; he built VoiceGremlin in roughly 45 days according to his launch post.
Founding Story
Nick Hall started VoiceGremlin after observing that automated testing against phone agents was technically doable but painful: telephony onboarding, audio/STT/TTS plumbing, LLM-to-LLM prompt tuning, and fraud-prevention work all create substantial implementation burden. The initial vision was to be a lightweight, vendor-agnostic 'Mailosaur for phone agents' that does one thing well, without managing test cases, hosting voice agents, or adding upsell products.
Business Model
Revenue Model
Monthly subscriptions billed through Stripe with an included allowance of call minutes; usage is charged by call duration rather than per test. There are no setup costs or overage charges, and unused minutes roll over up to three times the plan size.
Pricing Tiers
100 one-time minutes, 14-day trial, full transcripts, and API access; no credit card required.
600 minutes, concurrency of 3, full transcripts, and API access.
2,000 minutes, concurrency of 5, full transcripts, and API access.
8,000 minutes, concurrency of 10, full transcripts, and email support.
Target Markets
- Software engineering and SDET teams
- Companies deploying intelligent voice response/virtual assistant systems
- Teams integrating customer-support, airline, hotel, restaurant, or other phone agents into CI/CD
- Developers who need vendor-agnostic testing rather than a voice-agent hosting platform
- Regression and functional testing of AI phone agents after deployments
- CI/CD smoke and end-to-end checks for voice-agent behavior
- Testing disclosure of AI status and call recording
- Testing authentication, PII handling, transfers to humans, and other behavioral requirements
- Testing inbound voice agents without building an outbound test harness
- Optional open-ended adversarial or red-team scenarios, although the site says the product is not optimized for them