Opik is an open-source AI observability and evaluation platform built by Comet for debugging, evaluating, and monitoring LLM applications, RAG systems, and agentic workflows. The platform provides deep LLM call tracing with detailed context in both development and production environments, supporting over 40 million traces daily at scale. For evaluation and testing, Opik offers dataset and experiment management, LLM-as-a-judge metrics for hallucination detection, moderation, and RAG assessment, and PyTest CI/CD pipeline integration. Production monitoring features include real-time feedback score and token usage tracking, online evaluation rules with LLM-as-Judge metrics, and an agent optimizer with guardrails for continuous improvement. The platform integrates with 60+ frameworks including OpenAI, LangChain, Anthropic, CrewAI, LlamaIndex, AWS Bedrock, Groq, Mistral, Google ADK, Autogen, and Flowise AI. Deployment options range from Comet.com cloud hosting requiring no setup, self-hosted Docker Compose for local development, to Kubernetes/Helm for enterprise-scale deployments. The Python SDK offers simple decorator-based tracing with the @opik.track annotation. Version 1.7.0 introduced service profiles for flexible deployment scenarios and Kubernetes security improvements. Built primarily in Python, TypeScript, and Java, Opik is Apache-2.0 licensed with over 4,700 commits and an active contributor community.