LangGraph now supports deferred node execution, allowing you to create more sophisticated flows where a node only runs after all parallel branches have completed.

This unlocks a powerful new class of agent and workflow designs — ideal for coordinating complex, asynchronous logic.

Key Benefits

• Map-Reduce Flows — Fan-out, process independently, and fan-in results at the right moment

• Consensus-Based Decisions — Wait for input from multiple paths before continuing

• Multi-Agent Collaboration — Coordinate workflows across agents with different runtimes

Check out the docs to get started : https://langchain-ai.github.io/langgraph/concepts/server-mcp/#exposing-an-agent-as-mcp-tool