Judgeval

Judgeval is the listing for Judgment Labs, a company that sells what it calls a continuous improvement stack for AI agents. The platform is aimed at teams running agents in production who need to find out when those agents misbehave, understand why, and confirm that a fix works before shipping it. Judgment connects to production traces and uses its own agents to investigate them. Teams can ask questions in Slack, for example why an agent issued a full refund when only one item was damaged, and get an answer that cites the relevant trace spans along with how often the behavior occurred and which customers were affected. The platform searches for similar failures across sessions, narrows down root causes, and turns a recurring failure into a tracked behavior so the team is told when it happens again. Proposed fixes can be checked against real production cases through agent tests, such as a guard that verifies refunds are escalated to a supervisor before credit is issued. The product is built for engineering teams that ship customer facing agents. The company announced $32M in funding led by Lightspeed in May 2026. The vendor does not publish prices. Visitors are directed to book a demo or log in to the platform.