LangChain has been exploring a different approach to AI apps, focusing on ambient signals and user input only when necessary. This allows agents to save human attention for critical interactions. LangGraph is a new platform that makes these patterns easy to implement. An email assistant demonstrating key ambient agent patterns has been built, showcasing the potential of this design paradigm. The importance of human-in-the-loop interaction with agents has been highlighted, bringing benefits such as lowered stakes, mimicking human communication, and empowering long-term memory and learning. The Agent Inbox is a new UX for interacting with ambient agents, displaying all open lines of communication between humans and agents, making it easy to track outstanding actions. LangGraph provides built-in persistence layer, support for human-in-the-loop patterns, long-term memory, and cron jobs, making it an ideal platform for building ambient agents. An email assistant has been developed using LangGraph, available both as a hosted email agent and open-source project.