OpenTelemetry logo
Observability

OpenTelemetry

Vendor-neutral standard for traces, metrics and logs. OpenTelemetry is an open-source observability framework under the Cloud Native Computing Foundation. It defines APIs, SDKs and a wire protocol (OTLP) for collecting traces, metrics and logs, plus the Collector for processing and routing that data. Most observability vendors and open-source backends accept it, and it has semantic conventions for generative AI calls.

When it is the right choice

  • You want to instrument once and keep the freedom to change observability backends
  • You need traces that follow a request across many services and model calls
  • You want to track token usage, latency and errors of AI features alongside normal services
  • Not a backend by itself: you still need somewhere to store and view the data, such as Grafana or Datadog

How we use it

  • Instrument services and LLM calls with OpenTelemetry SDKs and auto-instrumentation
  • Run the OpenTelemetry Collector to filter, sample and route data to your chosen backend
  • Correlate AI traces with infrastructure metrics so slow answers can be traced to a cause
  • Use the collected data in regular reviews to find and fix the biggest issues first
Works well with

Often used together

GrafanaPrometheusDatadogLangChain
Services

Where it fits in our work

AI observabilityKnow what is wrong, and why, before your customers tell youLearn more »After launchInfrastructure that keeps getting cheaper, faster and safer after launchLearn more »
From the blog

Related articles

AI · October 6, 2026Observability for LLM ApplicationsWhat to measure in an LLM application: tokens, latency, cost per request, OpenTelemetry traces and the quality signals that show when answers get worse.Devops · October 8, 2024MicroservicesExplore the benefits and challenges of microservices architecture, including scalability, fault isolation, and best practices for DevOps and CI/CD integration.Devops · October 1, 2024Serverless MonitoringLearn the key metrics and tools for serverless monitoring and observability, including AWS CloudWatch and Datadog, to ensure optimal performance
Questions

OpenTelemetry: frequently asked

Does OpenTelemetry replace Prometheus or Datadog?

No. It collects and sends telemetry; Prometheus, Grafana's backends or Datadog store and display it. It lets you change those backends without re-instrumenting code.

Can OpenTelemetry trace LLM calls?

Yes. There are semantic conventions for generative AI and libraries that instrument common model SDKs, so prompts, token counts and latency appear in traces. We make sure sensitive prompt content is handled according to your policy.

Running OpenTelemetry, or deciding whether to? Tell us what you have; the first consultation is free.

Let's talk »