You can now send node inputs/outputs from LangGraph Studio directly to LangSmith datasets for testing! This means you can open LLM runs from LangGraph Studio in the LangSmith playground, making it easier to debug, visualize, and refine your agents.

This is an early step toward making Studio a full-fledged iteration hub for LangSmith users. For example, you can now experiment with prompt engineering and testing within a thread. In the future, we aim to let you apply changes made in the playground directly to your agent.

Learn more about LangGraph Studio: https://langchain-ai.github.io/langgraph/concepts/langgraph_studio/

Try LangSmith today: smith.langchain.com