Skip to content

Claude Platform adds OpenTelemetry trace export in public beta

Messages API requests and Managed Agents sessions can now stream standard OTel spans into Datadog, Grafana, Honeycomb and ten other observability tools.

Source: Anthropic

Anthropic has put OpenTelemetry trace export for the Claude Platform into public beta. The Messages API, Managed Agents sessions and the Claude SDKs can now emit standard OTel spans covering model requests, tool uses and agent turns, built on the OpenTelemetry GenAI semantic conventions so most observability platforms recognise them without extra setup.

SDK tracing needs no special permission and works with any cloud platform. API tracing needs an organisation admin to add a destination in the Claude Console, and only covers Claude Platform or Claude Platform on AWS. Each destination gets a content mode: metadata only by default, which exports IDs, timings, token counts and tool names, or full content, which adds the actual prompts and responses, capped at 0.75 MB per span unless you raise it (a full 1M-token window needs about 4 MB). Anthropic lists twelve launch partners, including Datadog, Grafana Cloud, Honeycomb, Sentry and LangSmith, plus support for any standard OTLP endpoint. Organisations with a HIPAA, FERPA or K-12 restriction only ever get metadata, regardless of the mode chosen.

traces.max_content_bytes = 4_000_000  # room for a full 1M-token trace

Why it matters: anyone running Claude agents in production has probably built ad hoc logging just to answer why a run failed or cost too much. This turns that into a standard export into whatever observability stack the team already runs, instead of another bespoke integration.

The caveat: span delivery is best effort, not guaranteed, so this isn’t a billing-accurate usage record on its own. Anthropic points to the separate Usage API for that.

Share

Related reading

ML for Web Devs

Tokens, context windows, and why your bill is what it is

The unit of billing for language models is not the request or the word. Once you understand tokens and how a context window fills up, the invoice stops being a surprise.

ML for Web Devs

RAG is just search with extra steps

Retrieval-augmented generation gets written up like an architecture. It is a search query, a string concatenation and one API call, and the search query is the part that decides whether it works.

ML for Web Devs

A prompt is a spec, so write it like one

Prompt engineering has very little to do with magic words. It is the same skill as writing a clear ticket for a capable contractor who will not ask you any questions.

More from The Wire

All briefs

GitHub Copilot retires six AI models today

Gemini 3.7 Flash, GPT-5.5, GPT-5.4, GPT-5.4 mini, GPT-5 mini and Grok 4.5 stop working across Copilot Chat, inline edits, agent mode and completions from today.

via GitHub