LangChain and Notion Unveil Major AI Agent Infrastructure Updates

AI agents are evolving from one-off conversations into workflows that can run for hours and operate across multiple tools, making tracking, debugging and collaboration critical infrastructure. LangChain's LangSmith focuses on observing agent behavior, while Notion aims to turn documents and databases into a shared workspace for teams and external agents. The moves reflect an industry shift from model capabilities toward deployment and governance.

On May 13, 2026, LangChain launched SmithDB, reporting a P50 trace-tree loading time of 92 milliseconds and a full-text search time of 400 milliseconds. SmithDB now handles 100% of data writes and trace-interface queries in the U.S. cloud. Notion unveiled its Developer Platform the same day and, on July 1, named Claude and Cursor as the first External Agents able to receive and execute tasks on shared boards. The External Agents API remains in private beta.