Cleanlab

Cleanlab is a runtime validation layer for AI agents and LLM apps. Its Codex product detects hallucinations, retrieval errors, and policy violations in real time, and pairs detection with human-in-the-loop remediation and knowledge-base fixes.

It runs as an independent layer that scores responses without changing the underlying stack, and integrates with OpenAI, Anthropic, Azure, AWS, and others. Available as SaaS or private VPC deployment.