Repolode
Submit
Arize-ai

Arize-ai/phoenix

Open-source, self-hosted AI observability platform for tracing, evaluating, and debugging LLM applications and agents.

Open-source alternative to:LangSmithHelicone

View on GitHubWebsite
Stars
11,907
Forks
1,219
Open issues
1,172
Last commit
yesterday

What is it?

Arize Phoenix is an open-source AI observability platform for experimenting with, evaluating, and troubleshooting LLM applications. It records runtime traces of LLM calls using OpenTelemetry-based instrumentation, so developers can see what their application did step by step.

On top of traces, Phoenix offers response and retrieval evaluations, versioned datasets, and experiments for comparing changes to prompts, models, and retrieval. It also includes a playground for replaying traced LLM calls, prompt management with version control and tagging, and a built-in agent for debugging traces. A remote MCP server lets coding agents such as Claude Code and Cursor query traces and experiments.

Phoenix can run locally with pip or uvx, or be deployed with Docker, Kubernetes via a Helm chart, or on several cloud platforms. It integrates with frameworks such as LangGraph, LlamaIndex, CrewAI, and DSPy, and with providers including OpenAI, Anthropic, and AWS Bedrock.

Key features

  • 01Traces LLM application runtime using OpenTelemetry-based instrumentation.
  • 02Runs response and retrieval evaluations on application outputs.
  • 03Manages versioned datasets and experiments for comparing prompts, models, and retrieval.
  • 04Provides a playground for replaying traced LLM calls and managing prompts.
  • 05Exposes a remote MCP server so coding agents can query traces and datasets.

Use cases

  • Debugging multi-step LLM agents by inspecting each traced call and tool use.
  • Comparing prompt or model changes against a fixed dataset before release.
  • Self-hosting LLM observability in a private environment on Docker or Kubernetes.

Quick start

  1. 1

    Install Phoenix with pip

    pip install arize-phoenix
  2. 2

    Start the Phoenix server

    phoenix serve

Things to know

  • License is listed as NOASSERTION, so the license terms are unclear from metadata and should be checked before use.
  • Full setup options and production deployment require reading the external install and self-hosting docs.

Topics

#llm-evaluation#observability#tracing#opentelemetry#llmops#prompt-engineering#evals#agents#ai-monitoring#ai-observability

FAQ

What is phoenix?
Open-source, self-hosted AI observability platform for tracing, evaluating, and debugging LLM applications and agents. Arize Phoenix is an open-source AI observability platform for experimenting with, evaluating, and troubleshooting LLM applications. It records runtime traces of LLM calls using OpenTelemetry-based instrumentation, so developers can see what their application did step by step.
Is phoenix free to use? What license does it have?
phoenix is open source under the Other license. Check the license terms before using it in commercial products.
Is phoenix still maintained?
Actively maintained. The last commit to phoenix was on Oct 19, 2026.
What can I use phoenix for?
Debugging multi-step LLM agents by inspecting each traced call and tool use. Comparing prompt or model changes against a fixed dataset before release. Self-hosting LLM observability in a private environment on Docker or Kubernetes.
How do I install phoenix?
Install Phoenix with pip: pip install arize-phoenix Start the Phoenix server: phoenix serve
Where can I get phoenix?
The source code and installation instructions are on GitHub: https://github.com/Arize-ai/phoenix

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects