Arize AX
Agent observability, evaluation and improvement platform for AI engineering teams.
- Tool type
- Agent observability and evaluation platform
- Made for
- AI engineers and AI product teams
- Runs on
- Web app with SaaS or self-hosted deployment
- Used in the Arize web app; SaaS and self-hosted deployment options are listed on the pricing page. Alyx opens with Cmd+L or Ctrl+L.
From @arizeai, about Arize AX
Recent important posts · launches, features
- Launch
Eval Health reviews evals and proposes fixes in Arize AX
“Who’s checking your evals? Eval Health is our new managed agent in Arize AX. It reviews your project’s evals and proposes changes to fill coverage gaps, fix broken checks, or remove ones that tell you nothing useful. It also spots LLM judges you could replace with cheaper code,”
@arizeai · Oct 28View - New feature
Trace Comparison with Diff Mode arrives in Arize AX
“Why are we debugging agent runs like diff was never invented? Trace Comparison in Arize AX puts two traces side by side, with Diff Mode highlighting changed values and added or removed spans. Try it in Arize AX:”
@arizeai · Oct 27View
What it does
Arize AX takes in traces from AI agents and applications, including spans, sessions and evaluation data, and produces observability dashboards, evaluation results, experiments and proposed fixes. Signal surfaces issues in production and suggests review-ready changes, while Alyx, the built-in AI engineering agent, runs evals, investigates errors and makes changes after asking for approval unless auto-accept is enabled.
Teams use it to trace LLM calls, compare prompt variants on datasets, attach online or offline evaluators to traces, and replay traces in a playground. It is instrumented with OpenInference and OpenTelemetry, and its adb datastore connects to BigQuery, Databricks or Snowflake. The open-source Phoenix project is a separate product built on the same standards, and it can be used alongside Arize AX.
It suits AI engineers building chatbots, RAG systems, copilots or agents. Access is through the web app, with a free plan of 25k spans and 1 GB per month and 15-day retention, and paid tiers for higher volumes. Enterprise pricing is quote-based. Alyx's model choice depends on which models have enough tool-use capability.
What it can do
- Trace LLM calls, agent trajectories and sessions with OpenTelemetry-compliant observability
- Run online evaluations on traces and spans and offline evaluations on datasets and experiments
- Custom code evaluators, agent-as-judge and multi-modal evaluation (image, voice, pdf)
- Prompt playgrounds, dataset management, prompt versioning and trace replay in playground
- Signal to detect production issues and propose fixes
- Alyx, an in-app AI engineering agent that works across traces, datasets, experiments and evaluators
Works with
- OpenAI
- Anthropic
- Amazon Bedrock
- LangGraph
- LangChain
- LlamaIndex
- CrewAI
- OpenAI Agents SDK
- DSPy
Example workflows
- Debug a failing agent from traces
- Input: traces from a project where a search tool errors. Steps: send traces via OpenInference instrumentation, ask Alyx to find traces where the search tool failed and group the error messages, then review proposed changes. Output: grouped error messages and an optional evaluator to catch the failure pattern.
- Compare prompt variants on a dataset
- Input: an existing dataset of prompts. Steps: ask Alyx to create two prompt variants and attach a helpfulness evaluation, then run the experiment in the playground. Output: side-by-side experiment results for the two variants.
Before you choose
- Alyx runs on an Arize-managed model by default; only models with enough tool-use capability appear in the model selector.
- The free plan includes 25k spans and 1 GB per month with 15-day retention and 10 Signal issues per month.
- Phoenix is the separate open-source product; Arize AX is the managed platform built on the same standards.
Listed under
Get started
Connect an agent or application to Arize and send a first trace.
- Create a free account at https://app.arize.com/auth/join.
- Connect your agent or application to Arize using OpenInference or OpenTelemetry instrumentation and send a first trace.
- Optionally add your own LLM provider under Settings → Account Settings → Integrations and pick a model in Alyx's model selector.
Plans and prices
Prices in USDAX Free is a free plan with no credit card required. AX Pro is listed at $50 per month. Enterprise pricing is custom and is requested through a demo. Alyx issue limits and any overage charges are not detailed on the pricing page.
- 25k spans per month
- 1 GB ingestion per month
- 50k spans per month
- 10 GB ingestion per month
Source: arize.com/pricing, checked 29 Oct 2026. Prices can change; check before you buy.
Sources for this overview
Product pages, documentation, and pricing checked 29 Oct 2026. Each source supports part of the overview; availability and prices can change.
Ratings and reviews
No ratings for Arize AX yet.
What’s changed
Free accountSignals
Free accountData and security
Free accountCompare with similar tools
Free accountWhere this comes from
How we research- Company website
- arize.com: product, pricing and policy pages
- YC directory mirror
- Batch and company name
- Social accounts
- Accounts the company links, and its posts on X
- Human review
- Checked by a person
- knowledge_base
Last checked 29 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.