LangChain Releases Managed Deep Agents v0.9 with Schedules, Per-Run Configuration, and Slack Reactions

LangChain has released Managed Deep Agents v0.9, introducing capabilities that allow agents to operate more autonomously within team environments like Slack. The update focuses on three main features: agent-created schedules, dynamic per-run configuration, and Slack message reactions.

Agent-Created Schedules

The new Schedules SDK allows agents to set up reminders, follow-ups, and recurring tasks directly from a conversation. When an agent creates a schedule, it runs with the permissions of the user who requested it, and results are posted back to the original channel. This supports both recurring cron jobs and one-time follow-ups, enabling requests like "Remind me about this tomorrow" to be handled natively by the agent.

Dynamic Per-Run Configuration

Agents can now be configured dynamically for each run, allowing a single deployment to serve multiple teams or use cases. By defining the agent as a callable function that receives runtime context, developers can select the specific model, instructions, skills, and MCP servers before the model executes. This approach prevents non-deterministic tool selection by the model and acts as a form of access control, ensuring agents only see the tools relevant to their current context.

Slack Reactions

To improve user experience, agents can now react to Slack messages immediately upon receipt, acknowledging the request while they process it. Reactions are enabled by default with the 👀 emoji, but developers can customize this behavior using a function to select different emojis based on message content, such as using 🐛 for messages mentioning broken features.