Arize AI

Arize AI is an AI engineering and observability platform for LLM applications, agents, and traditional ML systems. The commercial Arize AX platform (with Generative and ML & CV variants) provides tracing, evaluation, experiments, prompt management, and the Alyx AI engineering agent, built on the OpenInference OpenTelemetry conventions. Phoenix is the open-source counterpart used by tens of thousands of developers for local tracing, evaluation, and prompt iteration. Arize is vendor- and framework-agnostic with 30+ instrumentation providers and an OTLP-native ingestion path.

4 APIs 8 Features

LLM ObservabilityML MonitoringOpen SourceOpenTelemetryPhoenixTracingEvaluation

LLM Evaluation

Run built-in and custom evaluators on production traces, experiments, and datasets.

Experiments

Compare prompt and model variants over curated datasets with structured logging.

Prompt Management

Playground, hub, builder, and versioning for prompts used across applications.

Annotations

Capture human feedback on traces and outputs for evaluator development and dataset curation.

Alyx AI Engineer

AI assistant for debugging, evaluator authoring, dashboarding, and experiment comparison.

ML Monitoring

Drift, data quality, and performance monitoring for traditional ML and computer vision models.

Phoenix OSS

Open-source local tracing and evaluation tool runnable in notebooks or self-hosted.

Use Cases

LLM Application Observability

Monitor production LLM applications with traces, evaluators, and alerting.

Agent Debugging

Inspect multi-step agent runs across tool calls and intermediate reasoning.

RAG Quality Monitoring

Evaluate retrieval and generation quality over time in RAG systems.

ML Monitoring

Detect drift and degradation in classical ML and CV models.

Local Development

Iterate on prompts and evals locally with Phoenix before shipping to Arize AX.

Integrations

OpenAI

OpenInference instrumentation for OpenAI Chat Completions, Assistants, and Responses APIs.

Anthropic

Instrumentation for Anthropic Claude models.

LangChain

Instrumentation and evaluators for LangChain chains and agents.

LangGraph

Trace and evaluate LangGraph stateful agents.

LlamaIndex

Instrumentation for LlamaIndex RAG pipelines.

CrewAI

Trace CrewAI multi-agent crews.

DSPy

Trace and evaluate DSPy programs.

Vercel AI SDK

Instrumentation for Vercel AI SDK applications.

OpenTelemetry

OTLP-native ingestion compatible with any OTel collector or backend.

Bedrock

Instrumentation for AWS Bedrock model invocations.

Vertex AI

Instrumentation for Google Vertex AI and Gemini.

Resources

PhoenixDocumentation

GitHubOrganization

Sources

opencollection: 1.0.0
info:
  name: Arize AX OTLP Ingestion API
  version: 1.0.0
request:
  auth:
    type: apikey
    key: api_key
    value: '{{api_key}}'
    placement: header
items:
- info:
    name: Traces
    type: folder
  items:
  - info:
      name: Ingest OpenTelemetry traces (OTLP/HTTP)
      type: http
    http:
      method: POST
      url: https://otlp.arize.com/v1/traces
      headers:
      - name: space_id
        value: ''
      - name: api_key
        value: ''
      - name: Content-Type
        value: ''
      body:
        type: json
        data: '{}'
    docs: 'Accepts an OpenTelemetry ExportTraceServiceRequest payload encoded as

      either Protobuf (application/x-protobuf) or JSON (application/json). Use

      the space_id and api_key headers (or the OTLP standard authorization

      header carrying the same key) to authenticate.

      '
bundled: true

Arize AI

APIs

Arize AX

Phoenix

OpenInference

Alyx

Collections

Arize AX OTLP Ingestion API

Pricing Plans

Arize Ai Plans Pricing

Rate Limits

Arize Ai Rate Limits

FinOps

Arize Ai Finops

Features

Use Cases

Integrations

Resources

Sources