LangChain Launches Managed Deep Agents v0.9 in Beta

LangChain has released Managed Deep Agents v0.9 into public beta, bringing new capabilities designed to make AI agents function more like human teammates. A major addition is the Schedules SDK, which allows agents to generate their own reminders, follow-ups, and recurring tasks directly from an ongoing conversation. When a user asks an agent to handle a recurring task, the agent can programmatically schedule it using cron expressions or one-time at commands. These schedules run under the initiating user's permissions and automatically post results back to the original channel.

The update also introduces dynamic per-run configuration, allowing a single agent deployment to adapt its parameters for every individual execution. Instead of maintaining separate agent deployments for different teams or code repositories, developers can write a single callable function. This function inspects the runtime context to dynamically select the most appropriate model, instructions, skills, Model Context Protocol (MCP) servers, and sandboxes. For example, a coding agent can load a cheaper model and specific testing tools depending on the repository it is accessing, which reduces context window usage and prevents non-deterministic tool selection.

Additionally, version 0.9 improves user experience in chat applications with customizable Slack reactions. Agents can now instantly acknowledge incoming messages with emojis, such as a default eyes emoji or a bug emoji when a system failure is mentioned, while they process complex reasoning steps in the background. Developers can even use a fast, inexpensive decision model like Jev to select these emojis dynamically.

This release follows the recent v0.8 update, which introduced per-user memory, custom HTTP channels, and web search powered by Parallel. Developers can get started with the new beta by initializing a project with the command uvx --from managed-deepagents mda init my-agent and deploying it using uv run mda deploy. The system integrates with LangSmith to help developers debug agent decisions and manage deployments.