The Voice AI Gym
Where AI voice agents come to train. Test your AI against real IVR systems and AI agents that simulate humans.
Build deterministic phone trees or AI-powered responders. Track performance and success rates. Generate training data from every test call.
See It In Action
Design test scenarios. Call your AI agents. Review transcripts. Improve.
Real phone infrastructure. Real conversations. Real feedback.
How It Works
Get a Phone Number
We provision a real DID. Build your test scenarios using our visual dialplan builder or AI agents.
Call with Your Agent
Have your voice AI call the number. Test against IVR trees or AI agents simulating real humans.
Review & Export Training Data
Review transcripts, correct responses, and export to OpenAI for fine-tuning your custom model.
Built for Systematic Testing
Everything you need to test voice AI agents at scale and generate high-quality training data.
Deterministic IVR Scenarios
Build phone trees with predictable paths. Set success goals and track completion rates.
AI Agent Testing
Create AI responders that simulate real humans answering the phone. Test natural conversations.
Performance Benchmarking
Track success rates, measure goal completion, and benchmark across test runs.
Repeatable Test Cases
Save scenarios as "gyms" - reusable test suites you can run repeatedly.
Human-in-the-Loop Review
Review transcripts, correct agent responses, and build a gold-standard dataset.
Export Training Data
Export reviewed conversations directly to OpenAI for fine-tuning your custom models.
Ready to Test Your Agent?
Start testing and generating training data in minutes
No credit card required to create account
Already have an account?
Simple, Transparent Pricing
Perfect for testing your first AI agent
- 1 DID (U.S. or Canada)
- 1,000 minutes included
- Unlimited test scenarios
- Auto-transcription & review
- Export training data to OpenAI
For teams training multiple AI agents
- 3 DIDs included (add more for $10/mo each)
- 5,000 minutes included
- Priority support (Slack/Discord)
- Agent prompt versioning
- Performance metrics (latency, errors, TTFB)
For production AI agent deployments
- 10 DIDs included
- 15,000 minutes included
- All Team features
- Webhook exports & API access
- Fine-tuning dataset exports
- Audit logging & compliance

Where AI voice agents come to train. Built for builders.