Opik

Opik is an open‑source platform that provides end‑to‑end observability for generative AI applications, from prototype to production. It records detailed traces of LLM calls, conversation flows, and agent activity, allowing developers to monitor performance, track costs, and visualize the execution of complex systems. The platform also includes evaluation tools such as prompt testing, LLM‑as‑a‑judge, and experiment management, enabling systematic assessment of model outputs.

Built for developers, Opik offers a suite of features that support software‑testing‑style workflows for AI agents. Users can define regression tests and assertions, run end‑to‑end agent playground sessions, and automatically generate fixes that are written back to the codebase with accompanying test cases. These capabilities aim to reduce guesswork and streamline the iteration cycle for LLM‑driven products.

The project is self‑hostable, released under the Apache‑2.0 license, and provides a free tier without subscription requirements. It is positioned as a stable, production‑ready solution for teams that need deep tracing, online evaluation rules, and prompt optimization while maintaining open‑source transparency.