Features
Full trace logging for every agent step
30+ LLM-as-a-judge metrics including hallucination detection
Test Suites for simplified pass/fail evaluation
Ollie auto-fix coding agent that writes fixes to git
Opik Agent Playground for rapid agent prototyping
Cost intelligence for Claude Code and Codex spend
Human annotation and debugging of individual traces
Production monitoring for agents
Experiment management with custom visualizations
Model versioning and dataset management
@track decorator for easy integration with any LLM framework
Self-hosted or cloud deployment
Real-time trace streaming and visualization
Supports OpenAI, LangChain, LlamaIndex, Claude Code, Codex
Integration with PyTorch, TensorFlow, Hugging Face, and more
Trace visibility for agent steps (prompts, retrievals, tool calls, outputs)
LLM-as-judge evaluation for relevance, toxicity, quality scoring
Dataset creation from traces for reproducible testing
Experiment management and regression benchmarking
Built-in Prompt IDE for iterative prompt optimization
Self-hosted deployment on local, Docker, Kubernetes
Phoenix Cloud managed hosting option
Vendor-agnostic support for any model/framework
Native OpenTelemetry integration
OpenInference specification for LLM telemetry
Human annotation and automated labeling
Ghost trajectories to simulate alternative agent paths
Eval-as-you-test for early quality feedback
One-click integration with LlamaIndex