CheckOps
Arize Phoenix
Arize Phoenix is an open-source (Elastic License 2.0) AI observability and evaluation tool for tracing, evaluating and improving LLM apps and agents, with a hosted Arize AX tier.
- LLM & agent evaluation
- AI-enabled testing
- Observability
- Open source
- Free
- SaaS
- Paid
- Free tier
- Release & CI gates
- Production monitoring & observability
Latest public signals
Not refreshed yet. Refresh reads the public site and, when the entry has a GitHub repository, the latest release.
Introduction
Phoenix is Arize’s open-source AI observability and evaluation project. The site describes it as a way to trace, evaluate and improve AI agents, running locally or self-hosted. The repository Arize-ai/phoenix is licensed under the Elastic License 2.0 (ELv2), which is source-available rather than an OSI-approved licence.
Arize also sells Arize AX, a managed AI engineering platform with free, Pro and Enterprise tiers. Phoenix can be started locally with one command (uvx arize-phoenix serve).
Features
- Tracing: Capture traces from LLM apps and agents, based on OpenTelemetry instrumentation.
- Evaluations: Run LLM-as-judge and code evaluations on traces and datasets.
- Datasets and experiments: Compare versions of prompts and apps.
- Local or self-hosted: Run on a laptop or your own infrastructure.
- Coding agent setup: The site shows adding tracing with a coding agent through an
npxinstaller. - Managed AX: Hosted tiers for teams.
Pros
- Free to self-host: No licence fee for Phoenix.
- Quick start: One command to run locally.
- Popular: Over 12,000 GitHub stars at review.
- Path to managed: Move to Arize AX if needed.
- OpenTelemetry-based: Portable instrumentation.
Cons
- ELv2 licence: Restrictions apply, such as not offering it as a managed service; check with legal.
- Self-hosting effort: You run storage and upgrades.
- Judge costs: LLM evaluations cost tokens.
- Two products: Phoenix and AX differ in features.
- Observability-first: Testing workflows need setup.
Usage and Links (Example)
Start Phoenix locally, from the official site:
uvx arize-phoenix serve
Instrument your app with the Phoenix or OpenInference libraries, send traces, then create datasets and run evaluations in the UI or from Python.
Resources:
Pricing
Phoenix is free to self-host under ELv2. From Arize’s pricing page for the managed AX product:
- AX Free: Free. 25k trace spans a month, 1 GB ingestion, 15-day retention, unlimited users and evals.
- AX Pro: $50 a month. 50k spans, 10 GB, 30-day retention.
- AX Enterprise: Custom.
Recommended for?
- Teams wanting self-hosted LLM tracing and evals at no licence cost.
- Developers who want a local UI for debugging agents.
- Organisations planning to grow into managed observability.
- Not a fit if you require an OSI-approved open-source licence.
Conclusion
Phoenix is a popular, easy-to-start tool for tracing and evaluating LLM apps, with a managed path through Arize AX.
Read the ELv2 terms before embedding or hosting it for others.
Alternatives
Other tools on CheckOps that cover similar ground. Open a page, or compare one with Arize Phoenix side by side.
LLangfuseLangfuse is an open-source platform for LLM traces, prompts, and evals. Self-hosting is free; Langfuse Cloud is paid after a hobby tier.Compare with Arize Phoenix
LLangSmithLangSmith is LangChain's platform for observability and evaluation of LLM applications and agents, including offline and online evals.Compare with Arize Phoenix
OOpikOpik by Comet is an Apache-2.0 open-source platform to trace, evaluate and monitor LLM apps, RAG systems and agents, with test suites, LLM-as-judge evals and a free cloud tier.Compare with Arize Phoenix
BBraintrustBraintrust is a hosted platform for AI evals and agent observability: trace production, turn patterns into evals, score releases, and query it from your IDE via MCP.Compare with Arize Phoenix
TTruLensTruLens is an MIT-licensed, OpenTelemetry-native library for tracing and evaluating LLM apps and agents with explainable LLM judges, maintained by Snowflake.Compare with Arize Phoenix
RRagasRagas is an open-source library for systematically evaluating LLM applications, including RAG systems and agents.Compare with Arize Phoenix