Opik is an open-source platform for evaluating, testing, and observing LLM applications and AI agents. Developed by Comet, it provides tracing and experiment-management tools that help teams understand prompts, model calls, tool use, latency, costs, and output quality across development and production workflows.

The platform supports LLM traces, datasets, experiments, automated and custom evaluators, prompt comparison, and integrations with popular model and agent frameworks. Teams can use it to debug individual executions, run repeatable evaluations before releases, and monitor how an application behaves as models, prompts, or retrieval logic change.

Opik is useful for developers who need a systematic quality loop around generative-AI software rather than relying on manual spot checks. Its open-source deployment option also makes it relevant to organizations that want to keep evaluation and observability data inside their own infrastructure.