Arize AI (Phoenix / AX)
AI observability and evaluation: open source Phoenix for local tracing/evals plus hosted Arize AX for production.
Best for Teams needing evals plus production LLM monitoring
Used Arize AI and liked it? Upvote it. One upvote per person. Upvotes come from visitors, not from us, and never change "Our ranking".
What it does
Arize AI is an agent observability and evaluation platform available as open source Phoenix and hosted Arize AX. Users connect AI agents and applications to capture execution traces and run evaluations to inspect behavior and improve performance. It is built for AI engineers and teams developing, monitoring, and scaling production generative AI systems.
Read more: 5 features and how people use themShow less
-
Application and agent tracing
Captures execution traces across spans, LLM calls, and agent steps to show what occurred inside each request. Uses open standards so teams can instrument once without proprietary trace formats.
Use caseA staff engineer at Wayfair investigates failures during pre launch development and conducts post launch debugging on agent requests.
From arize.com · image from Arize AI -
Comprehensive evaluation framework
Executes span, trace, and session evaluations across models and workflows at scale. Operates locally or in real time online to quantify whether system changes improve output quality.
Use caseA principal engineer at PagerDuty runs real time online evaluations to track performance while developing agents.
From arize.com -
Alyx AI engineering agent
Investigates reported issues, executes evaluations, and produces review ready fixes for agent architectures. Operates directly inside AI native workflows alongside tools like Cursor and Claude Code.
Use caseAn AI engineer assigns a failing agent issue to Alyx to uncover the root cause and generate a fix.
From arize.com -
Phoenix open source platform
Runs locally or self hosted to trace every model call and execute evaluations while keeping data private. Serves as a local first tool for experimentation and prompt iteration prior to production rollout.
Use caseA software developer traces LLM calls and tests prompts locally on their machine without sending data to a cloud host.
From arize.com/pricing -
Automated stack instrumentation
Detects project frameworks and provider wrappers via a single terminal command to wire up tracing integrations. Connects with dozens of major LLM providers and agent libraries across Python and TypeScript.
Use caseA developer runs a terminal command to detect their framework and configure telemetry before signing up for Arize AX.
From arize.com/docs/ax/integrations · image from Arize AI
Written from Arize AI's own website, checked . Each feature links to the page it comes from; images are Arize AI's own.
Every plan
| Plan | Monthly | Billed yearly | Per | What it includes |
|---|---|---|---|---|
| Free (AX Free) | Free | Not listed | per account | 25k spans/mo, 1 GB ingestion/mo, 15 day retention, unlimited users, SaaS only |
| AX Pro | $50 | Not listed | per account | 50k spans/mo, 10 GB ingestion/mo, 30 day retention, 2 spaces, email support |
| AX Enterprise | Not listed | Not listed | per account | contact sales; SaaS or self hosted, custom retention, SLA |
| Phoenix (open source) | Free | Not listed | per account | self host free, Elastic License 2.0 |
From Arize AI (Phoenix / AX)'s own pricing page, checked . Prices are in US dollars per month; "billed yearly" is the monthly price when you pay for a year up front. A few figures (often the yearly price) are derived rather than printed on the page.
Before you pay
- Card needed for the trial
- No
- HIPAA
- Yes
What Arize AI says
Arize supports flexible deployment options and is certified to leading security and compliance standards, including SOC 2 Type II, ISO 27001, PCI DSS, HIPAA, and GDPR.
Source page - SOC 2
- Yes
- GDPR
- Yes
What Arize AI says
Arize supports flexible deployment options and is certified to leading security and compliance standards, including SOC 2 Type II, ISO 27001, PCI DSS, HIPAA, and GDPR.
Source page - Discounts
- Startup pricing is available upon application.
What Arize AI says
Apply for startup pricing to trace, evaluate, and iterate on your LLM applications.
Source page - Support
- Community (Slack), email, dedicated
Only what Arize AI's own pages state, with the quote behind each answer. Anything they don't publish is left out rather than guessed.
Getting started with Arize AI
- Run the setup command from your project root (e.g. npx evals, uvx arize ax cli instrument, or ax instrument) to create your account and API key.
- Select which coding agent to use and choose whether to trace an existing app, a new starter app, or your coding agent.
- Add your LLM provider's API key, such as OPENAI_API_KEY, into your .env file.
- Run your app until the first trace arrives.
- Open your project to view trace details, examine each span tree, and check token counts, latency, and status.
Written from Arize AI's own help pages: arize.com/docs/ax/get-started/get-started-tracing
Arize AI video tutorials
Independent creators on YouTube, not Arize AI or us. The video loads from YouTube only when you press play.
What the free plan includes
AX Free: 25k spans/
Questions people ask about Arize AI
How much does Arize AI cost?
Paid plans start at $50 a month (AX Pro). 4 plans are listed above, from Arize AI's own pricing page.
Is Arize AI free?
Yes, there is a free plan: AX Free: 25k spans/mo, 1 GB ingestion, 15 day retention, 1 space; Phoenix self hosted free.
Is Arize AI HIPAA compliant?
Yes. Source
Do I need a credit card to try Arize AI?
No, you can start without a credit card. Source
Does Arize AI have a student or other discount?
Startup pricing is available upon application. Source
Does Arize AI have an API?
Yes.
How do I get started?
Connect your agent or application to Arize and send your first trace. Traces show what happened inside each request, giving you the data you need to debug, evaluate, and improve your agents. Source
Does Arize work with my stack?
Arize integrates with 40+ models, frameworks, and AI tools, including OpenAI, Anthropic, Google, Amazon Bedrock, LangGraph, LangChain, LlamaIndex, CrewAI, OpenAI Agents SDK, DSPy, and more. Source
What about data privacy?
Your data stays under your control. Arize supports flexible deployment options and is certified to leading security and compliance standards, including SOC 2 Type II, ISO 27001, PCI DSS, HIPAA, and GDPR. Source
What’s the difference between Arize products?
Phoenix is the open source platform for AI observability and evaluation that can be run locally or self hosted, whereas Arize AX is the managed AI engineering platform providing managed infrastructure, adb datastore, advanced observability, online evals, and continual improvement workflows. Source
Where this came from
Plans, prices and limits were copied from Arize AI (Phoenix / AX)'s own pages and the other sources listed, then checked .
- arize.com/pricing/
- arize.com/docs/phoenix/self-hosting/license
- arize.com/
- arize.com/docs/ax/get-started/get-started-tracing
- arize.com/docs/ax/integrations
- arize.com/trust-center/
- arize.com/privacy-policy/
- arize.com/docs/ax/alyx
- arize.com/docs/ax/set-up-with-ai-assistants
How we collect prices and rank tools · Is something wrong? Correct this listing
Others that do the same job
Braintrustbraintrust.dev
Evaluation first AI platform for experiments, scoring, prompt playgrounds and production tracing of LLM apps.
- Green flag: Free plan
- Green flag: No card for trial
- Red flag: Credits reset monthly
LangSmithlangchain.com/langsmith
LangChain's platform for tracing, evaluating, and deploying LLM apps and agents, framework agnostic.
- Green flag: Free plan
- Green flag: Doesn't train on your data
Langfuselangfuse.com
Open source LLM engineering platform for tracing, evals, prompt management and cost tracking, cloud or self hosted.
- Green flag: Free plan
- Red flag: Acquired
Heliconehelicone.ai
Open source LLM observability and gateway: proxy based request logging, cost tracking, caching and prompt management.
- Green flag: Free plan
- Green flag: No card for trial
- Red flag: Acquired
Weights & Biases Weavewandb.ai/site/weave
W&B toolkit for tracing, evaluating and iterating on LLM apps and agents, now under CoreWeave Forge.
- Green flag: Free plan
- Red flag: Acquired
Datadog LLM Observability (Agent Observability)datadoghq.com/product/llm-observability
Datadog module tracing LLM apps and agents end to end with quality, safety and cost monitoring.
- Green flag: Free plan
- Red flag: Renamed
Free plan · Flat monthly
Visit site Datadog LLM Observability (Agent Observability) (opens the tool's site)