Opik
Open-source LLM observability, evaluation, and agent tracing — debug, test, and monitor the full lifecycle.
About
Opik is Comet's open-source platform for the full LLM application lifecycle: trace every LLM call with full trace trees for multi-step agents and tool calls, run evaluations with datasets and LLM-as-a-judge metrics (hallucination detection, moderation, RAG assessment), manage prompts in a playground, and monitor production with dashboards and online evaluation rules. A PyTest integration lets you test LLM pipelines on every commit, and the Agent Optimizer SDK tunes prompts and agents automatically.
It integrates natively with the frameworks teams actually use — Google ADK, AutoGen, Flowise, LangChain, and more — and it is Apache-2.0 with a genuinely self-hostable full platform (Docker Compose), not just an open client. At 22,600+ stars it is the leading open alternative to LangSmith-style tracing. The business model is the usual open-core shape: the platform is free to self-host, and Comet's cloud is where the upsell lives, so evaluate the managed tier on its own merits.
What it can do
- Full trace trees for LLM calls, agents, and tool use in dev and production
- Evaluation: datasets, experiments, and LLM-as-a-judge metrics
- Prompt playground and the Agent Optimizer SDK for tuning
- Production monitoring dashboards with online evaluation rules
- Native integrations: Google ADK, AutoGen, Flowise, LangChain, and more
- PyTest integration for CI/CD testing of LLM pipelines
Setup
- Self-host with Docker Compose, or sign up for Opik Cloud (free tier).
- Install the SDK:
pip install opik(TypeScript SDK also available). - Initialize in your code and wrap LLM calls — traces appear in the dashboard.
- Build datasets and run experiments with LLM-as-a-judge metrics.
- Add the PyTest integration to CI and set production monitoring rules.
Prerequisites
Python or TypeScript; a self-hosted Opik instance (Docker) or a Comet cloud account.
Use cases
Works with
Trust score — 9/10
Scored 9/10 (verified tier, metrics-v1). Evidence: 22,626 GitHub stars (GitHub API), Apache-2.0, pushed Oct 15 2026; the README documents tracing, evaluation, the optimizer SDK, guardrails, and the framework integrations; docs corroborate the self-host and CI flows. Strengths: the leading open LLM-observability platform, genuinely self-hostable, Comet backing. Docked for: (1) open-core upsell to Comet Cloud; (2) crowded eval/observability category. License: Apache-2.0.
★ 22,626 GitHub stars
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 8.0/10
- Maintenance · recency, releases, issue load
- 10.0/10
- Openness · license + docs
- 10.0/10
- Manual review · sanity read + red-flag check
- 9.0/10
- · Adoption 8: starScore from 22,626 stars on comet-ml/opik
- · Maintenance 10: pushed 2026-10-15 (recency 10); frequent releases (10); 187 open issues on 22.6k stars (issues 8)
- · Openness 10: Apache-2.0 (0.7*10) + official docs verified at comet.com/docs/opik (0.3*10)
- · Manual 9: README documents tracing, evaluation, optimizer SDK, guardrails, and framework integrations; docked for open-core upsell to Comet Cloud and crowded category
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.