Arize Phoenix is Arize's open-source AI observability platform designed for experimentation, evaluation, and troubleshooting. For managed production workflows, Arize also offers Arize AX. Phoenix provides:
Tracing - Trace your LLM application's runtime using OpenTelemetry-based instrumentation.
Evaluation - Leverage LLMs to benchmark your application's performance using response and retrieval evals.
Datasets - Create versioned datasets of examples for experimentation, evaluation, and fine-tuning.
Experiments - Track and evaluate changes to prompts, LLMs, and retrieval.
Prompt Management- Manage and test prompt changes systematically using version control, tagging, and experimentation.
PXI (Phoenix Intelligence) - An AI engineering agent built into Phoenix for debugging traces, iterating on prompts, and navigating the product.
Remote MCP Server - Connect Claude Code, Cursor, and other MCP clients directly to your Phoenix instance's /mcp endpoint to query traces, datasets, experiments, and more.
Phoenix runs practically anywhere, including your local machine, a containerized deployment, or in the cloud. See Environments for a walkthrough of each option, or jump straight into the Tracing Quickstart.
[!NOTE]
Phoenix is open-source and self-hosted. For a managed platform, Arize offers Arize AX — the same OpenTelemetry/OpenInference instrumentation works with both. See the full comparison.
The fastest way to send traces is to let your coding agent (Claude Code, Codex, Cursor, and others) instrument your app. From your project directory, run:
npx @arizeai/phoenix-cli setup
# or, with Phoenix installed: px setup
Setup detects your framework and LLM provider, installs the right OpenInference instrumentation, and wires up trace export. Prefer to wire it up in code? See the tracing documentation.
Deploy
Phoenix container images are available via Docker Hub and can be deployed using Docker or Kubernetes via the Helm chart.
[!NOTE]
The Google Cloud button builds Phoenix from source in Cloud Shell rather than deploying the prebuilt Docker Hub image. The Azure template serves plain HTTP (Azure Container Instances does not terminate TLS) — front it with a TLS proxy such as an Application Gateway before production use.
Packages
The arize-phoenix package includes the entire Phoenix platform. However, if you have deployed the Phoenix platform, there are lightweight Python sub-packages and TypeScript packages that can be used in conjunction with the platform.
CLI for fetching traces, datasets, and experiments for use with Claude Code, Cursor, and other coding agents
Tracing Integrations
Phoenix is built on top of OpenTelemetry and is vendor, language, and framework agnostic. For details about tracing integrations and example applications, see the OpenInference project and the integrations documentation.
Phoenix is built to be driven by people and by AI coding agents alike. Three surfaces let agents (Claude Code, Codex, Cursor, and others) work with your traces, datasets, and experiments:
CLI — npx @arizeai/phoenix-cli fetches traces, datasets, and experiments and instruments your app (setup), so an agent can pull context and act on it from the terminal.
Skills — .agents/skills/ packages workflows that teach agents how to debug, evaluate, and trace with Phoenix.
Remote MCP Server — connect any MCP client to your instance's /mcp endpoint to query Phoenix directly.
By default, Phoenix collects basic web analytics (e.g., page views, UI interactions) to help us understand how Phoenix is used and improve the product. None of your trace data, evaluation results, or any sensitive information is ever collected.
You can opt-out of telemetry by setting the environment variable: PHOENIX_TELEMETRY_ENABLED=false
Community
Join our community to connect with thousands of AI builders.
Arize Phoenix 是一个开源的 AI 可观测性平台，用于追踪、评估和排查 LLM 应用、智能体以及 RAG 流水线中的问题。