AWS Introduces CloudWatch Omni for AI Agent and Application Observability

Amazon CloudWatch now offers CloudWatch Omni as a unified observability experience for application and AI workloads, with an app-centric design, AI-powered features, open-standards foundations, and access outside the AWS Management Console. [2][3] CloudWatch Omni is purpose-built for observing, evaluating, and experimenting with AI agents across model providers, frameworks, and runtimes. [2] The service provides a native extension for VS Code and Kiro, while operators can use a standalone web experience accessed through single sign-on. [2]

CloudWatch Omni captures every trace and includes evaluators for correctness, coherence, retrieval quality, and tool selection. [2] Teams can compare prompt versions side by side, build test datasets from production traffic, run experiments across configurations, and automatically detect regressions. [2] Built on OpenTelemetry, Omni displays telemetry already sent to CloudWatch without reconfiguration and accepts other instrumented workloads through an OpenTelemetry Protocol endpoint. [3] Amazon says the product combines agent and application observability in one experience organized around applications rather than individual signals. [3]

CloudWatch Omni puts agent evaluation and application telemetry in the same workspace, reducing the split between testing AI behavior and diagnosing the systems around it. Compared with traditional monitoring, its built-in evaluators and trace capture target quality failures that may not appear as standard errors. Support for OpenTelemetry, VS Code, Kiro, and a standalone SSO web experience also gives teams several paths into the same observability layer rather than tying work to the AWS Management Console.

The practical test will be whether this unified workflow can replace the fragmented reviews and tool-switching AWS describes. What the sources don't address: pricing, regional availability, data-retention controls, and how Omni performs at production scale.