Arize Phoenix
A source-available (Elastic License 2.0) LLM observability and evaluation tool built on OpenTelemetry, runnable anywhere.
Arize Phoenix is a source-available (ELv2) LLM tracing and evaluation tool built on OpenTelemetry/OpenInference. Run it locally in a notebook or self-host it to capture traces, run evals (including LLM-as-judge), and debug RAG and agent runs without sending data to a vendor.
Arize Phoenix is a source-available observability and evaluation tool for LLM applications. Built on OpenTelemetry and the OpenInference tracing standard, it captures the full trace of a run and lets you evaluate outputs — and because it runs locally or self-hosted, your traces never have to leave your environment.
It is aimed at engineers who want vendor-neutral observability they can spin up in a notebook during development and self-host in production. Phoenix is the free, self-hostable companion to Arize's commercial platform, so you can start free and graduate to the managed product if you outgrow it.
Highlights
- OpenTelemetry-native tracing — instrument with open standards (OpenInference), avoiding lock-in to one vendor's SDK.
- Run anywhere — launch locally in a notebook for dev, or self-host for team/production use.
- Built-in evals — LLM-as-judge and other evaluators for relevance, hallucination, and RAG quality.
- RAG & agent debugging — inspect retrieval steps, tool calls, and the full span tree behind an answer.
- Framework-agnostic — works across common LLM and orchestration stacks via auto-instrumentation.
In an AI-assisted workflow
phoenix serve # local UI for traces + evals
# or, with no install:
uvx arize-phoenix serve
# then auto-instrument your LLM/agent calls, inspect spans, and run evaluatorsTIP
Because Phoenix speaks OpenTelemetry, the instrumentation you add is portable — you can ship the same traces to another OTel-compatible backend later without re-instrumenting.
Good to know
Phoenix is free to self-host, but it is source-available rather than OSI open source: the Elastic License 2.0 bars offering it to third parties as a hosted or managed service, even though Arize's README still calls Phoenix open source. You bring an LLM provider for judge-based evals. Arize also offers a managed platform for teams that want hosted scale and support. For an open-source (MIT core) option with a managed cloud and first-class self-hosting, compare Langfuse; for the commercial LangChain-native option, LangSmith.
Frequently asked questions
- What is Arize Phoenix?
- Arize Phoenix is a source-available observability and evaluation tool for LLM applications (Elastic License 2.0), built on OpenTelemetry and the OpenInference tracing standard. It captures the full trace of a run, runs evals (including LLM-as-judge) for relevance, hallucination, and RAG quality, and lets you debug RAG and agent runs by inspecting the span tree behind an answer.
- Is Arize Phoenix free?
- Yes — Phoenix is free to self-host under the Elastic License 2.0 (source-available; it bars offering Phoenix to others as a hosted service), and because it runs locally or self-hosted, your traces never have to leave your environment. You bring an LLM provider for judge-based evals, and Arize offers a managed commercial platform if you outgrow it.
- How do I use Arize Phoenix?
- Start it locally with phoenix serve (or uvx arize-phoenix serve, with no install), auto-instrument your LLM or agent calls, then inspect spans and run evaluators in the local UI. Because Phoenix speaks OpenTelemetry, the instrumentation is portable — you can ship the same traces to another OTel-compatible backend later without re-instrumenting.
- Arize Phoenix vs Langfuse?
- Both are free to self-host. Phoenix is source-available (ELv2), OpenTelemetry-native, and built to run anywhere — in a notebook during development or self-hosted in production — while Langfuse is open source (MIT core) with a managed cloud and first-class self-hosting; LangSmith is the commercial LangChain-native alternative.
Filed under
observability · tracing · evals · source-available · opentelemetry
Related
- LangfuseAn open-source LLM engineering platform for tracing, evals, prompt management, and metrics.
- LangSmithLangChain's platform for tracing, evaluating, and monitoring LLM apps — framework-agnostic.
- Best LLM & RAG Evaluation Tools in 2026: DeepEval vs RAGAS vs LangSmith vs Phoenix vs promptfooA decision guide to the LLM eval landscape — code-first frameworks vs. eval-and-observability platforms, open-source vs. hosted, and which fits your stack.
- LLM Observability EngineerUse this agent to make a production LLM app observable — tracing every step, scoring live traffic with online evals, and monitoring quality, cost, and latency — so you can debug agent runs and catch regressions in production. Examples — "add tracing to our RAG/agent so we can debug bad answers", "set up online evals and cost/latency dashboards", "production quality is slipping and we're flying blind".
- The Best LLM Observability Tools in 2026: Langfuse vs LangSmith vs Braintrust vs PhoenixLangfuse, LangSmith, Braintrust, Arize Phoenix, Helicone and AgentOps compared on tracing, cost tracking, agent debugging, OpenTelemetry and self-hosting.
- LLM Observability in Production: Traces, Evals, Cost, and FeedbackInstrument production LLM applications with end-to-end traces, online evals, cost and latency metrics, privacy controls, alerts, and a failure-feedback loop.
- AgentOpsObservability for AI agents — session replay, cost and latency tracking, and debugging for multi-step runs.