Opik
Open-source LLM tracing, evals and guardrails you can self-host
Overview
Opik is Comet's open-source platform for tracing, evaluating and guarding LLM and agent applications, with the same codebase running self-hosted or as managed cloud. It integrates with 50-plus providers and frameworks and has passed 20,000 GitHub stars.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 12 Jul 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
- Full observability and agent-testing feature set
- Agent tracing, test suites, playground
- No usage caps to manage yourself
- Up to 10 team members
- 25k spans/month
- 60-day data retention
- Up to 50 team members
- 100k spans/month, +$5 per extra 100k
- Customizable span limits and retention
- Unlimited team members
- SSO, service accounts, view-only users
- SOC 2, ISO 27001/9001, HIPAA, GDPR
What you can produce with Opik
- Distributed tracing for LLM, RAG and agent apps
- LLM-as-a-judge plus heuristic evaluators
- Bayesian prompt and agent optimizer
- Real-time guardrails for PII and policy violations
- Self-hostable via Docker or Kubernetes Helm chart
- 50+ provider and framework integrations
- Production dashboards and online evaluation rules
ASEAN Perspective
Opik in Southeast Asia
ASEAN-region availability and pricing notes coming soon. Drop the editorial team a note via /contact/ if you can supply local context (Singapore/Malaysia/Indonesia/Thailand/Vietnam).
Opik is the rare open-source LLM observability tool where the free self-hosted version isn't a stripped-down teaser — Comet's own docs describe it as the same codebase as the hosted product. Tracing, LLM-as-a-judge evaluation, a prompt/agent optimizer, and a guardrails layer for PII and policy violations all ship in the open-source core.
Comet built its reputation on experiment tracking for classic ML, and that dashboard polish carries over; production dashboards and evaluation rules are more mature than most OSS alternatives. Self-hosting via Docker is quick, but the Kubernetes/Helm path for production has real setup-order dependencies — ClickHouse replication has to come up after Opik, not before — worth budgeting time for. At 20,000-plus GitHub stars it has real adoption momentum. A solid default pick for teams that want LangSmith-grade tooling without SaaS-only lock-in.
What people say
Opik comes out of Comet ML, the experiment-tracking company that's served classic ML teams since well before the LLM wave, and it shows in the finish: the tracing and evaluation dashboards feel like a mature analytics product rather than a weekend open-source project. The repository crossed roughly 18,300 GitHub stars by March 2026 and has since grown past 20,000 — fast for an 18-month-old project in a crowded category.
The feature set is broad for something free: distributed tracing, LLM-as-a-judge scoring (Comet's own benchmarking claims 85-92% correlation with human evaluators on hallucination and relevance checks), a Bayesian prompt optimizer that early adopters have used to cut inference cost 20-40%, and a guardrails layer that screens for PII and policy violations in real time. All of it ships in the same open-source codebase Comet runs in its own cloud, rather than a cut-down community edition.
On setup: local Docker Compose deployment is genuinely quick — reviewers report a working instance in under 10 minutes — though Comet is explicit that Compose isn't meant for production load. The Kubernetes Helm chart is the recommended production path and is described internally as battle-tested, but has ordering gotchas (ClickHouse replication must be configured after Opik is already running, not before) that have tripped up self-hosters in community discussions.
G2 reviews of Comet's broader platform, which Opik sits inside commercially, praise dashboard depth and experiment-tracking maturity, with the recurring complaint being that per-seat pricing scales awkwardly for larger teams and the UI can lag on very large projects. Pricing for Opik specifically starts free (25k spans/month hosted, 60-day retention, up to 10 members), with a $19/month Pro Cloud tier at 100k spans and $5 per additional 100k — cheap relative to Braintrust or LangSmith at similar volume. Enterprise is custom and adds SSO plus SOC 2, ISO 27001, HIPAA and GDPR coverage.
Summary of public user & expert reviews, compiled by RECATOOLS.
About this listing
This entry was compiled from publicly available data including Opik's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with Opik unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to Opik directly →
Spotted something out of date? Suggest an update →
Opik in the news
Developer Tools
Miasma: How a Hijacked CI Pipeline Shipped Malicious npm Packages With Valid Provenance
Developer Tools
OpenTelemetry Graduates CNCF: 12,000 Contributors Lock In the Industry's Unified Observabi...
Developer Tools
GitLab 19.0 drops Redis for Valkey by default and brings Gitaly to Kubernetes
More in Code & Dev Tools
