OpenTelemetry on AWS
Send OpenTelemetry metrics, logs, and traces from your AI agents and applications to Amazon CloudWatch, with no proprietary agent required.
What is OpenTelemetry?
Give your AI agents full context. OpenTelemetry, the open source framework hosted by the Cloud Native Computing Foundation, puts metrics, logs, and traces from every service and agent onto the same data standard. Amazon CloudWatch is OpenTelemery native end to end, so observability data is stored, queried, and shared with full context to other agents.
Interoperable and vendor neutral
OpenTelemetry is developed by a broad community of contributors, including AWS, and no single vendor controls the specification or its roadmap. Amazon Cloudwatch’s native OpenTelemetry compatibility lets you connect to other agents and tools without re-instrumentation.
Built for every major language
OpenTelemetry APIs and SDKs are available for languages including Java, Python, Go, .NET, and JavaScript, with zero-code instrumentation for popular libraries and frameworks.
Adopted across the industry
Most observability tools and cloud providers run on OpenTelemetry data, which has made it the default method teams use to instrument new applications.
Benefits
Use the OpenTelemetry SDKs, or collect telemetry from supported runtimes without changing your application. Your instrumentation stays the same as your tooling changes, so you can easily swapwith a configuration change.
Metrics, logs, and traces share one data model and naming conventions, so you can move from an alarm to the log event to the trace behind it. Your dashboards, automations, and AI agents operate in the same context across every service.
Use your existing OpenTelemetry Collector or start with CloudWatch Agent. Collect from sources such as Prometheus, collectd, StatsD, and JMX and send that telemetry data to Amazon CloudWatch with no application changes.
OpenTelemetry data powers CloudWatch Omni
Amazon CloudWatch Omni is the AI-first experience in CloudWatch that investigates issues across your metrics, logs, and traces. It reasons over the OpenTelemetry conventions you adopt, so when you send telemetry over OTLP, your workloads appear in Omni with no additional configuration.
Use cases
Replace third-party agents and classic custom metrics with OpenTelemetry, and pay for each GB you ingest instead of for each metric and each alarm. Send to your current tool and CloudWatch in parallel during the migration, then use PromQL to alarm on thousands of resources in a single query.
Get cluster, node, pod, and container telemetry from Amazon Elastic Kubernetes Service (Amazon EKS) and Amazon Elastic Container Service (Amazon ECS), with cluster and workload metadata attached to every signal.
Send telemetry from on-premises hosts and other cloud providers to Amazon CloudWatch over OpenTelemetry Protocol (OTLP), using the same instrumentation and collector configuration as your AWS workloads.
Capture model invocations, token usage, and tool calls as traces using the OpenTelemetry semantic conventions for generative AI. View latency, errors, and cost for each request alongside the rest of your application.