Coval

Coval: Simulation and Evaluation for AI Voice and Chat Agents

Coval is a testing and observability platform for teams shipping AI voice and chat agents. It borrows the simulation-first discipline used to validate autonomous vehicles: instead of discovering failures on real customer calls, you run thousands of realistic simulated conversations against your agent before launch, then keep monitoring live traffic for regressions and quality drift after it ships.

The platform spans the full agent lifecycle. Pre-deployment, Coval simulates callers with varied personas, accents, and goals. In production, it observes real calls and flags failures. A human review layer routes conversations to QA reviewers and converts their feedback into reusable evaluation standards, so your test suite improves as your agent does.

Key Features

• Simulation engine that runs thousands of test conversations against voice and chat agents before each release
• Production observability that monitors live calls and detects failures, anomalies, and quality drift
• Human review queues that turn QA feedback into structured, repeatable evaluation criteria
• Behavior-level evals for identity verification, escalation handling, information collection, hallucination detection, and topic adherence
• SOC 2 Type II, HIPAA, and GDPR compliance included on all plans
• Published pricing: Starter at $100/month, Growth at $500/month, Enterprise from $4,500/month, each with a 7-day free trial

Ideal Use Case

Coval fits engineering and QA teams putting voice or chat agents into production for support, scheduling, collections, or healthcare workflows — anywhere an agent failure has real cost. It is especially useful for teams that need regression testing wired into their release process rather than ad-hoc manual call reviews.

How Coval differentiates

Founder Brooke Hopkins previously led evaluation job infrastructure at Waymo, and Coval applies that same simulation methodology to conversational agents. Founded in 2024 and a Y Combinator graduate, the company closed a $28M Series A led by Norwest in June 2026 (with Base10 Partners, Twilio Ventures, and Y Combinator), bringing total funding to $31M. Its site lists customers including Zoom, Perplexity, ServiceNow, StubHub, Toast, GEICO, and Hippocratic AI.

FAQ

Is there a free trial? Yes. All plans, including Starter and Growth, offer a 7-day free trial; a credit card is required and you can cancel any time.

Does Coval only test voice agents? No. It evaluates both voice and chat agents, covering text and audio conversation flows.

What does Coval cost? Starter is $100/month, Growth is $500/month, and Enterprise plans start around $4,500/month with custom limits.

Is Coval compliant for regulated industries? The platform ships with SOC 2 Type II, HIPAA, and GDPR compliance on every tier, which is why healthcare and financial customers appear on its client list.

tl;dr

Coval brings Waymo-style simulation testing to AI voice and chat agents — simulate before launch, monitor in production, and close the loop with human review. Backed by a $28M Norwest-led Series A and used by Zoom, GEICO, and Perplexity.

Related

Looking for more options? Browse the Developer Tools directory or read our best AI coding tools listicle. Coval is also tracked on Crunchbase.