Opik is an open-source platform for tracing, testing and monitoring apps built on language models, including RAG systems and AI agents.
In plain words
When an AI app gives a bad answer, you need to see what happened inside it. Opik logs each model call, tool call and agent step as a trace you can inspect. You can turn bad traces into test cases, score outputs with automated checks, including a model acting as judge, and watch quality, latency and cost once the app is live. It is Apache-2.0 licensed, and you can run the whole platform on your own servers.
A quick example
Imagine a support agent that sometimes gives wrong refund rules. In the trace you see it fetched an outdated document. You save that trace as a test case, fix the retrieval step, and rerun the test suite to confirm the answer is now right.
Where you'll meet it
- Debugging multi-step agents and RAG pipelines.
- Automated checks on LLM apps in CI on every commit.
- Monitoring quality and cost of a live AI app.
Related
Sources
- Docscomet-ml/opik, Comet (GitHub) · read 29 Sept 2026
- DocsOpik documentation, Comet · read 29 Sept 2026
Checked by WhatIsAI against the sources above · 29 Sept 2026. Spotted an error? Report it.