Roark

Roark: Voice AI Testing & Evals Platform

Roark is a quality assurance and evaluation platform designed for voice and chat AI teams. Rather than building agents, Roark focuses on testing, monitoring, and improving existing voice AI agents by simulating calls before launch and scoring production calls using audio-native metrics.

Key Features

• Simulation Testing: Run agents against hundreds of simulated callers, including realistic personas, red teaming for prompt injections, and 45 languages with native accents and background noise.
• Post-Call Analysis: Score every production call using over 500 audio-native and conversational metrics, measuring pronunciation, emotion, interruptions, empathy, and hallucination directly from the waveform.
• Self-Improving Loop: Features a prompt optimizer, ground-truth tuning, and fix verification to draft edits from failing calls and replay fixes against the exact callers that broke them.
• Load and Health Tests: Conduct peak-volume concurrency testing and always-on health checks to catch outages before customers do.
• Enterprise Security: Includes SOC 2 Type II compliance, HIPAA BAA availability, annual pen tests, SSO/SAML, and role-based access.

Use Cases

• Healthcare: Evaluate calls for required disclosures and identity checks automatically.
• Customer Support and Sales: Score production voice AI on pronunciation, empathy, and resolution across conversations.
• Pre-Launch Validation: Validate client voice agents in staging environments with regression testing and CI/CD gates before they go live.

Pricing

Users can start for free with a $50 credit, requiring no credit card. Comprehensive pricing tier information is not available on the website.

Getting Started

Website: https://roark.ai

Roark integrates with platforms like Vapi, Retell, LiveKit, and Pipecat, and offers SDKs for Node and Python alongside a REST API.