Anthropic says Claude Platform can export OpenTelemetry traces from Managed Agents, the Messages API and SDKs in public beta.
Anthropic said the Claude Platform now supports OpenTelemetry tracing for Claude Managed Agents, the Messages API and Claude SDKs, giving developers a way to inspect model calls, tool use and MCP server calls in standard observability tools. The company said the feature is available in public beta and is meant to help teams review agent behavior before widening access.
In material published on the Claude Resources site on Wednesday, 14 October 2026, Anthropic said developers can add a destination in the Claude Console and send every agent turn, model call and tool use to an observability stack as an OpenTelemetry trace, with latency, token counts and IDs included. The company said Claude traces follow the OpenTelemetry GenAI semantic conventions, which should allow data from Claude and other providers to be viewed in the same tools and dashboards.
The release notes published the same day on the Claude Platform also describe OpenTelemetry export for Messages API requests and Managed Agents sessions as a beta release. According to the platform documentation, the Claude Platform includes the Claude API, client SDKs and Claude Console, while the Messages API supports conversational interactions and the Managed Agents product handles programmatic agent workflows.
Anthropic said the tracing output is compatible with Grafana Cloud, Honeycomb, Sentry and Splunk, where agent traces can sit alongside logs and metrics. The company also named Arize AI, Braintrust, Datadog, Langfuse, LangSmith by LangChain, New Relic and PostHog as tools that can use the traces to turn production data into evals and measure whether changes improve results.
Source link







