Arize Phoenix

What is Arize Phoenix?

Arize Phoenix is an open-source observability and evaluation platform for LLM applications and AI agents. Your app sends traces over OpenTelemetry, and Phoenix shows each step of a run: model calls, retrieval, tool use and custom logic. You can then score outputs with LLM-based evaluators, code checks or human labels, group traces into datasets, and rerun them through new prompt versions to see whether a change helped. It has built-in instrumentation for frameworks such as LlamaIndex, LangChain, DSPy and the Vercel AI SDK, and for OpenAI, Bedrock and Anthropic, in Python, TypeScript and Java. Phoenix runs on a laptop, in Docker or on Kubernetes with Helm, and traces can stay inside your own network. Arize AI, the company behind it, also sells Arize AX, a managed platform with a free tier and a $50 per month Pro plan. Phoenix itself is free to self-host with no feature limits, so most teams start there and add AX only when they want a hosted service.

Use cases

• Trace every step of an LLM agent run
• Score model outputs with LLM judges
• Build datasets from production traces
• Compare prompt versions on the same inputs
• Replay LLM calls with different inputs
• Self-host observability in an air-gapped network

Who it's for

• AI engineers
• LLM application developers
• ML platform teams
• Teams with strict data-privacy needs

Why use Arize Phoenix

• See why an agent answered the way it did, because each prompt, retrieval and tool call appears in the trace.
• Turn quality into numbers with evaluators and datasets, so a prompt change is judged on evidence instead of a hunch.
• Keep traces and prompts on your own infrastructure, including fully air-gapped setups, with nothing sent to Arize.
• Reuse OpenTelemetry instrumentation, so the same traces can later go to Arize AX or other tools without re-instrumenting.
• Start for free on a laptop with a single command and move to Docker or Kubernetes when the team grows.

Arize Phoenix features

• Tracing over OpenTelemetry (OTLP) with OpenInference instrumentation
• Auto-instrumentation for LlamaIndex, LangChain, DSPy, Mastra and Vercel AI SDK
• LLM-based, code-based and human-annotation evaluations
• Evaluator integrations for Ragas, Deepeval and Cleanlab
• Datasets built from traces or uploaded from code and CSV
• Experiments that compare application versions on the same inputs
• Prompt management with versioning, a side-by-side playground and span replay
• Docker and Helm deployment, with OAuth2, LDAP and role-based access controls
• PXI, an agent inside Phoenix for debugging traces and iterating on prompts

Arize Phoenix integrations

• LlamaIndex
• LangChain
• DSPy
• Mastra
• Vercel AI SDK
• OpenAI
• Anthropic
• Ragas
• Deepeval
• Cleanlab
• OpenTelemetry

How to get started with Arize Phoenix

1. Start Phoenix locally with uvx arize-phoenix serve, or deploy it with Docker or the Helm chart.
2. Add OpenInference auto-instrumentation for your framework or model provider so traces flow to Phoenix over OpenTelemetry.
3. Open a trace to inspect prompts, retrievals, tool calls and outputs, and label good and bad runs by hand or with an LLM judge.
4. Save traces as a dataset, change a prompt or model, and run an experiment on the same inputs.
5. Compare evaluation scores, cost and latency, and keep only the changes that improve quality.

Arize Phoenix limitations

• Self-hosting means your team provisions, scales and maintains the deployment
• Support is community-based; SLAs and enterprise support come with Arize AX
• Licensed under ELv2 rather than a permissive open-source license

Arize Phoenix pricing

Arize Phoenix is open source, so you can use and self-host it for free.

Arize Phoenix plans

• Phoenix (self-hosted)

  Free

  open source, you host it

  • No license fees or usage limits
  • Tracing, evals, datasets and prompts
  • Docker, Helm or local install

• AX Free

  $0

  • 25k trace spans per month
  • 1 GB ingestion per month
  • 15 days retention
  • Unlimited users and evals

• AX Pro

  $50/mo

  • 50k trace spans per month
  • 10 GB ingestion per month
  • 30 days retention
  • Unlimited users and evals

• AX Enterprise

  Custom

  • Custom span volume and retention
  • SaaS or self-hosted deployment
  • Enterprise SSO and audit logs

Plans as listed on the official pricing page on 24 Oct 2026. Prices can change and may vary by country or exclude tax. Check the official pricing page.

Arize Phoenix alternatives

The main Arize Phoenix alternatives listed on Debuly are Langfuse, Helicone, Braintrust, Opik, Future AGI and VoltAgent. The table compares pricing and platforms; order is A–Z.

1. 

   Braintrust

   AI observability and evaluation platform for tracing LLM apps and agents, scoring outputs and catching regressions before release.

   Developer platformsCoding
   FreemiumVisit Braintrust (official site)
2. 

   Future AGI

   Platform for tracing, evaluating, simulating and guarding LLM apps and AI agents, with a gateway, usage-based billing and a free tier.

   Developer platformsAutomation and agents
   FreemiumVisit Future AGI (official site)
3. 

   Helicone

   Open-source LLM observability and AI gateway that logs requests, tracks cost and latency, and routes to 100+ models through one API.

   Developer platformsCoding
   FreemiumVisit Helicone (official site)
4. 

   Langfuse

   Open-source LLM observability and evaluation platform for tracing agents, managing prompts and scoring outputs.

   Developer platformsAutomation and agentsCoding
   FreemiumVisit Langfuse (official site)
5. 

   Opik

   Comet's open-source platform for tracing, evaluating and monitoring LLM applications and agents, self-hosted or in the cloud.

   Developer platformsAutomation and agentsCoding
   FreemiumVisit Opik (official site)
6. 

   VoltAgent

   Open-source TypeScript framework for building AI agents, paired with the VoltOps console for tracing, evals and deployment.

   Developer platformsAutomation and agentsCoding
   FreemiumVisit VoltAgent (official site)

Arize Phoenix FAQ

Sources & verification

Facts on this page come from the official sources below, each with the date we read it. Written by Debuly; the vendor did not review this page.

Image credits

• Logo: from arize.com · fetched 23 Oct 2026 · Vendor trademark, used only to identify the product; removed on request
• Cover: none collected yet, so a generated tile is shown.

Spot an error? Report an issue with this listing. We also remove a logo or image on the owner's request.