braintrust-tracing

What problem does it solve?

This Skill provides a comprehensive guide to integrating Claude Code sessions with Braintrust for detailed tracing, enabling effective debugging and analysis of agent interactions.

Core Features & Use Cases

• End-to-end Tracing: Understand the flow of execution from session start to end, including tool usage and LLM calls.
• Sub-agent Correlation: Learn strategies to link traces of parent sessions with their spawned sub-agents, even when they create orphaned traces.
• Debugging Tools: Utilize provided commands and logs to troubleshoot tracing issues and verify Braintrust integration.
• Use Case: When a complex multi-agent workflow fails, use Braintrust tracing to pinpoint the exact step and sub-agent responsible for the error by correlating parent and child session traces.

Quick Start

To enable Braintrust tracing, ensure TRACE_TO_BRAINTRUST is set to true in your .claude/settings.local.json and your BRAINTRUST_API_KEY is configured.