AI Potluck
Product / UX / Telemetry & observability

Opik

Comet ML

Platform for debugging, evaluating and monitoring LLM applications, RAG systems and agentic workflows, providing tracing, automated evaluations and production dashboards. Built by Comet ML, its server, web application and evaluation components all run on infrastructure you operate, with Comet's cloud as optional managed hosting.

Verified 2026-08-13 via the comet-ml/opik repository and the opik PyPI project page.

Openness

5 high confidence
5.0
license
Apache-2.0(OSI)
source
public
commercial
Comet-cloud-hosting(not a gated core)
core-gated
ungated

Apache-2.0, full platform self-hostable; cloud is managed hosting rather than a feature-gated proprietary core.

  • https://pypi.org/project/opik/ recorded 2026-08-13

    opik 2.2.28 on PyPI. The long description states "Apache-2.0 licensed, free to self-host the full platform", and its FAQ answers "Is Opik open source?" with "Opik is licensed under Apache 2.0. Its server, web application, and core observability and evaluation components" are covered.

  • https://github.com/comet-ml/opik recorded 2026-08-13

    Repository README, carrying the same statement - "Apache-2.0 licensed, free to self-host the full platform" - alongside the tracing, LLM-as-a-judge, dataset and integration sections. 21,353 stars.

Adoption

4 high confidence
4.0

opik draws 3,512,727 PyPI downloads a month, inside the 1M-10M band. It ships 60-70+ framework integrations, and roughly 19,400 GitHub stars corroborate without setting the band. The 40M+ traces a day the vendor once advertised is no longer on the repository README, so the band rests on the download count.

Capability

4 high confidence
4.0

Covers every dimension this category grades - OpenTelemetry-native tracing across 70+ integrations, datasets and experiments with pytest-driven CI, LLM-as-judge metrics for hallucination, moderation, relevance and context precision, a prompt playground, and cost tracking - which puts it level with the MLflow and Langfuse platforms that anchor 4. It is not scored 5 because the closed Arize AX leads the category on capability.

  • https://github.com/comet-ml/opik recorded 2026-08-13

    README sections - "Track all LLM calls and traces with detailed context during development and in production", LLM-as-a-judge scoring, "Annotate traces and spans with feedback scores", datasets and "Extensive 3rd-party integrations for easy observability".

Verified 2026-08-13