Arize AI

Arize AI is an enterprise observability, evaluation, and improvement platform for AI agents, designed to trace and optimize agent behavior in production at scale. Its architecture processes 1 trillion spans monthly and executes 1 billion evaluations monthly to support the largest AI deployments. The platform provides Arize AX (managed cloud version with complete infrastructure), Phoenix OSS (open-source tool for local tracing and evaluations with data control), and Alyx (an AI engineer agent that automatically debugs issues and suggests improvements). The product's core is the ability to trace end-to-end every agent call via OpenInference standards, execute LLM-as-judge or custom script evaluations, and deploy fixes confidently via integrated test/deploy workflows. Pricing: Free (25k spans/month, 15d retention), Pro ($50/month, 50k spans, 30d retention, unlimited users), Enterprise (custom with SaaS or self-hosted).

Arize does not provide a French-language interface and remains oriented toward English-speaking technical audiences. The API and SDKs (Python, TypeScript, Go, Ruby, C#) enable integration into any existing LLM stack via OpenInference standards, without framework dependency. Data can be stored in multiple cloud regions and the platform supports compliance requirements (SOC 2 Type II, HIPAA, GDPR). The platform excels for organizations with large-scale AI agent observability and evaluation needs, but requires significant expertise to configure evaluation pipelines and deploy workflows. The freemium model (generous Free tier) enables quick starts, but costs escalate rapidly if your span volumes exceed thresholds.