Automated agent triage with Agent Tracing and Claude Routines

A daily Claude Routine uses the Sentry MCP to analyze hundreds of Seer agent conversations and thousands of tool calls from the prior night, automatically identifying errors, unusual costs or latency, excessive tool use, and potential reasoning failures before the team begins work. Replacing a manual process that required extracting and parsing raw Langfuse trace data, the routine gathers aggregate metrics, samples both failed and successful conversations, examines underlying spans and source code, produces a report, checks Linear for existing issues, and files tickets for newly discovered problems. Its combination of aggregate error analysis and detailed conversation review revealed recurring cases in which the agent guessed incorrect repository or project identifiers, as well as backend latency issues, enabling targeted fixes such as supplying repository names directly. The experience suggests that reviewing successful runs alongside failures is essential for validating agent conclusions, while delegating data retrieval to Sentry and judgment to Claude creates a more consistent and efficient monitoring workflow.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.