The short version
- Amazon CloudWatch Omni is an AI-powered observability experience for applications and AI agents.
- AWS says Omni is built on OpenTelemetry and can bring telemetry from AWS accounts, regions and other cloud workloads into one space.
- The service can automatically discover services and dependencies and surface request rate, errors and duration metrics.
Amazon CloudWatch Omni is an AI-powered observability experience for applications and AI agents. Amazon CloudWatch Omni is an AI-powered observability experience for applications and AI agents. AWS says Omni is built on OpenTelemetry and can bring telemetry from AWS accounts, regions and other cloud workloads into a unified environment.
Observability for the agent era
Omni adds an agent observability workflow for frameworks including LangGraph, CrewAI, OpenAI Agents SDK, Vercel AI SDK and Strands.
Amazon CloudWatch documentation describes the observability features for AI applications and agents including support for several agent frameworks.
The service can automatically discover services and dependencies and surface request rate, errors and duration metrics. AWS is also adding an agent observability workflow for frameworks including LangGraph, CrewAI, OpenAI Agents SDK, Vercel AI SDK and Strands.
- AWS says Omni is built on OpenTelemetry and can bring telemetry from AWS accounts, regions and other cloud workloads into a unified environment.
- Omni adds an agent observability workflow for frameworks including LangGraph, CrewAI, OpenAI Agents SDK, Vercel AI SDK and Strands.
- AWS is also adding an agent observability workflow for frameworks including LangGraph, CrewAI, OpenAI Agents SDK, Vercel AI SDK and Strands.
Following the chain of models and tools
Open standards such as OpenTelemetry are important because enterprises rarely run one cloud, one model provider or one application framework.
CloudWatch Omni is aimed at a problem that becomes harder as applications span multiple cloud environments and include AI agents. Traditional monitoring can show latency, errors and resource use, but an agent may also generate a chain of model and tool calls that determines whether a task succeeds.
A common telemetry layer based on OpenTelemetry can make that information easier to collect across environments. Instead of maintaining completely separate monitoring systems for each cloud account or region, teams can bring telemetry into one operational view.
AI agents create a monitoring problem that traditional application dashboards do not completely solve. A task may involve several model calls, tool calls and intermediate decisions before the final result appears. An engineer therefore needs to see the sequence of events, not just the latency or error count of one service.
The agent observability features cover frameworks including LangGraph, CrewAI, OpenAI Agents SDK, Vercel AI SDK and Strands. That makes the developer tooling important because engineers need to connect application behavior with infrastructure signals.
The availability of a standalone web experience and IDE extensions for tools such as VS Code, Cursor and Kiro also points toward a developer-oriented workflow. The useful measure will be whether engineers can move from an unexplained agent failure to a specific trace, dependency or model interaction quickly enough to fix it.