Debugs LangChain and LangGraph agents by fetching and analyzing execution traces directly from LangSmith Studio.
LangSmith Fetch is a specialized debugging skill designed to streamline the analysis of LangChain and LangGraph agents. It allows developers to retrieve recent execution traces, investigate tool calls, check memory operations, and analyze performance metrics without leaving the terminal. By integrating directly with LangSmith Studio via the CLI, it provides actionable insights into agent failures, token usage, and complex execution flows, making it an essential tool for developers building and optimizing production-grade AI agent workflows.
Key Features
01Debug Long-Term Memory (LTM) and state operations
02Fetch and analyze recent execution traces
03Deep dive into specific tool calls and results
0422,944 GitHub stars
05Export full debug sessions for team collaboration
06Monitor token usage and execution latency
Use Cases
01Optimizing LLM costs by reviewing token usage per trace
02Identifying the root cause of agent execution errors
03Analyzing why an agent selected a specific tool or reasoning path
