Opik

What it does

Opik is an open-source platform for observing, evaluating, and improving LLM applications and agents. It logs every LLM call, tool invocation, and agent step so you can inspect the full chain of events behind any output, with token and cost tracking alongside.

Evaluation uses LLM-as-a-judge and heuristic metrics, plus test suites with pass/fail assertions that can be built from failing traces. Ollie, an AI assistant, can analyze traces and propose fixes to a connected codebase, the Agent Playground runs agents with every execution traced, and prompts can be versioned and optimized with several algorithms.

It suits developers and teams building chatbots, RAG pipelines, and multi-step agents. The open-source code can be self-hosted at no software cost. The hosted Free plan includes 25k spans per month, Pro is listed at $19 per month for 100k spans, and Enterprise pricing is custom.

What it can do

• Trace LLM calls, tool calls, and agent steps with execution graphs and session tracking
• Run automated LLM-as-a-judge and heuristic evaluations with 30+ built-in metrics and custom metrics
• Build test suites with pass/fail assertions and run them as unit or regression tests
• Version and optimize prompts with multiple optimization algorithms
• Apply PII, topic, and custom guardrails to LLM inputs and outputs
• Monitor production traces with online evaluation and alerts

Works with

• LangChain
• OpenAI
• Google ADK
• LangGraph
• CrewAI
• OpenTelemetry
• Ragas
• Hugging Face Datasets
• Gretel
• Model Context Protocol (MCP) server

Before you choose

• Pro plan is listed at $19 per month for 100k spans; Ollie tokens are purchased separately after the trial.
• Pro Cloud lists 60-day span data retention; Free Cloud also lists 60-day retention, with longer retention available as a paid add-on on Pro.
• Self-hosting requires you to provide and operate your own infrastructure.

Listed under

• Coding
• DevOps and observability
• Agent tracing
• Developers
• Evaluation
• LLM observability
• LLM testing
• MLOps

Get started

Free account requires no credit card; open-source code is available on GitHub with a self-hosting guide.

1. Install the Opik SDK and add a few lines of code to start logging traces from your application.
2. Turn failing traces into test cases and build a test suite to verify fixes.
3. Optionally install the Opik MCP server to manage workspace data from your AI IDE or chat tool.

Pro is listed at $19 per month for 100k spans; additional spans are $5 per 100k and extended retention is $29 per 100k spans. Ollie is free to trial, then tokens are purchased. Enterprise pricing requires contacting sales.

Source: comet.com/site/pricing, checked 29 Oct 2026. Prices can change; check before you buy.

Sources for this overview

Product pages, documentation, and pricing checked 29 Oct 2026. Each source supports part of the overview; availability and prices can change.

• Open-Source AI Observability Platform | Opik by Comet ↗www.comet.com
• Comet | Flexible Pricing Plans, Free for Academics ↗www.comet.com
• Opik Documentation - Open-Source LLM Observability & Optimization | Opik Documentation ↗www.comet.com

Company website

comet.com: product, pricing and policy pages

YC directory mirror

Batch and company name

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

knowledge_base

manual_review

Last checked 29 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.