Arize Phoenix is an open-source tool for tracing and evaluating AI apps, built by Arize AI, that you can run on your own machine or servers.
In plain words
Phoenix shows what happened during each run of an AI app, step by step: model calls, retrieval, tool use and custom logic. It collects traces through OpenTelemetry, a common standard, and plugs into popular frameworks and model providers. You can score traces with LLM judges, code checks or human labels, and test prompt changes on datasets before rolling them out.
A quick example
Imagine a RAG app that gives vague answers. The developer runs Phoenix locally, turns on tracing for their framework, and sees that the retrieved passages are off-topic. They add an LLM-based relevance evaluator and track the score as they fix the retriever.
Where you'll meet it
- Debugging agents and RAG pipelines.
- Self-hosted LLM observability.
Related
Sources
- DocsWhat is Arize Phoenix?, Arize AI documentation · read 29 Sept 2026
- DocsArize-ai/phoenix, Arize AI (GitHub) · read 29 Sept 2026
Checked by WhatIsAI against the sources above · 29 Sept 2026. Spotted an error? Report it.