Coval is a simulation and evaluation platform for AI agents, founded in 2024 by Brooke Hopkins, who previously led the evaluation job infrastructure team at Waymo. The company graduated from Y Combinator's Summer 2024 batch and raised a $3.3 million seed round led by MaC Venture Capital with participation from YC and General Catalyst. The core idea is to bring the testing discipline that autonomous vehicle teams developed over a decade—simulation-driven testing, continuous benchmarking, trajectory-based analysis—to voice and chat AI agent development. Coval's platform lets engineering teams test thousands of simulated conversation flows, including phone-call-style voice scenarios with realistic audio environments, so they can catch regressions before they reach production users. The platform generates diverse test cases automatically from a small set of seed cases, simulating edge cases and unexpected user behaviors without requiring engineers to hand-write each scenario. Evaluations run against built-in and custom metrics (accuracy, latency, instruction compliance, tool-call correctness) as well as full workflow validation. Production monitoring is also built in: all live calls are logged and continuously scored so teams get early warning of quality degradation. Coval targets engineering teams building voice-first or chat-first AI agents—IVR replacements, customer-support bots, scheduling assistants—who need structured, repeatable quality gates beyond manual spot-checking. The platform is SOC 2 Type II, HIPAA, and GDPR compliant across its tiers, making it viable for regulated industries such as healthcare and finance. As of 2026, Coval operates from its main domain coval.ai (with coval.dev also active) and publishes an independent TTS benchmark at benchmarks.coval.ai/tts, reinforcing its positioning as an objective evaluation infrastructure provider rather than an agent builder. Key features: - Automated simulation of thousands of conversation flows from a small set of seed test cases - Voice-channel testing including phone-call simulation with customizable voices and acoustic environments - Built-in and custom metrics: accuracy, latency, instruction compliance, tool-call validation - Continuous production monitoring with live call logging and early failure detection - Permutation and load testing to stress-test agent behavior at scale - SOC 2 Type II, HIPAA, and GDPR compliance across all tiers - Independent TTS benchmark (benchmarks.coval.ai/tts) for evaluating text-to-speech providers
Starter tier starts at $100/month (100 simulation minutes, 1K monitored calls); Growth tier at higher volume (1,000 simulation minutes, 10K calls); Enterprise tier with custom limits. 7-day free trial available. Historical tiers (late 2024) were Core $300/mo and Scale $999/mo.