# Claude Platform adds native OpenTelemetry export to Databricks MLflow

published: 2026-11-03T21:05:30.978Z
topic: infrastructure
source: https://www.databricks.com/blog/claude-platform-now-supports-native-opentelemetry-observability-databricks-mlflow
url: https://aiagentnews.io/stories/claude-platform-adds-native-opentelemetry-export-to-databricks-mlflow

Claude Platform now sends native OpenTelemetry traces from Anthropic SDKs, the Messages API and Claude Managed Agents to Databricks MLflow, in Public Beta. The traces are stored in Unity Catalog tables linked to MLflow experiments.

Native OpenTelemetry export covers the Anthropic SDKs, the Messages API and Claude Managed Agents. Claude's instrumentation creates spans for steps such as model requests and tool calls and exports them over the OpenTelemetry Protocol. Databricks receives them through Zerobus Ingest and writes them to Unity Catalog tables linked to an MLflow experiment.

Setup has three steps: create an experiment with Unity Catalog trace storage, configure the export with the workspace ID, region and spans table, then run an agent and open its trace. Claude Code and Claude Agent SDK can send traces to the same destination through additional environment variables. Inputs and outputs are captured only when content mode is set to full.

Once traces arrive, MLflow offers trace analysis, MLflow Issues in Beta for recurring problems across traces, Review Queues for expert feedback, and Agent Evaluation to compare agent versions on the same cases.

## Why it matters if you build agents

Agent developers can inspect Claude model and tool spans alongside traces from other instrumented services and reuse the same judges to compare a baseline agent with a revised one. Unity Catalog controls access to the stored trace data, which can be sensitive. Check the content mode setting before enabling full capture in production.