ToolsNews
Claude Platform adds OpenTelemetry traces for agents
Anthropic launched OpenTelemetry trace export on the Claude Platform in public beta. Calls from Claude Managed Agents, the Messages API and Claude SDKs now show up as standard traces in your own observability tool.
By: Mikola Máté Patrik · · 2 min read

What is new
Per Anthropic's October 14 announcement, every agent turn, model call and tool use can be sent as a standard OpenTelemetry trace, with latency, tokens and IDs (Anthropic). MCP server calls are traced too. If you already send Claude Code or Cowork traces to your observability tool, agents built on the Claude Platform will show up in the same place.
How it works
Managed Agents and the Messages API need a destination configured in the Claude Console. Claude SDKs export traces automatically. Any tool that accepts a standard OTLP endpoint works.
Traces follow the OpenTelemetry GenAI semantic conventions, so data from Claude and other providers fits in the same dashboards.
Which tools
Anthropic highlights three uses. Debugging in Grafana Cloud, Honeycomb, Sentry and Splunk, where agent traces sit beside logs and metrics. Improvement and evals with Arize AI, Braintrust, Datadog, Langfuse, LangSmith, New Relic and PostHog. Organisation-wide token tracking via Databricks Unity Gateway.
Among the quoted customers, Ramp says its own system that scores Langfuse traces has cut token usage by 10–20% on some agents and made sessions 30% faster. That is a customer claim, not an independent measurement.
Why it matters
Anthropic's argument is simple: widen an agent's access once its traces show it is reliable. Until now that often meant custom instrumentation. If you already run an OTel stack, this is close to free observability for Claude agents, and it gets easier to see which step was slow or expensive.
Produced with the help of AI, reviewed by an editor. How we make our articles
Reviewed and edited by the Neural Daily desk. About


