Introducing Amazon CloudWatch Omni: AI-powered observability for generative AI and agentic workloads

This article announces Amazon CloudWatch Omni, a unified observability solution for AI agents and generative AI workloads that combines development and operations capabilities.

• Native IDE extensions for VS Code and Kiro with traces, playground, and evaluators integrated directly into development environment
• Standalone web experience for operators to monitor agent fleets without AWS Management Console access
• 17 built-in evaluators for quality metrics including coherence, helpfulness, faithfulness, and routing correctness
• Supports popular frameworks: LangChain, LangGraph, CrewAI, OpenAI SDK, Vercel AI SDK, and Amazon Bedrock AgentCore
• Uses open standards (OpenInference and ADOT) for framework-agnostic instrumentation across Lambda, ECS, EKS
• Includes Trace Explorer with compare mode, Playground for prompt testing, Experiments for configuration comparison, and Session Explorer for multi-turn interactions
• Optional Cloud Login connects local development to AWS for persistent storage, team sharing, and production monitoring
• IDE extension is free; only AWS credentials needed for Bedrock models or API keys for other providers

CloudWatch Omni addresses observability challenges for non-deterministic AI agent behavior by providing comprehensive tracing, evaluation, and experimentation capabilities across development and production environments.