Multi-agent Swarms with LangGraph and Workato

Langchain has launched a Python library that enables the development of swarm-style multi-agent systems using LangGraph. This architecture empowers agents to dynamically transfer (hand off) control among themselves, leveraging their unique specializations. Notably, the system retains memory of the last active agent, ensuring seamless conversation continuity by resuming with the same agent in subsequent interactions.

GitHub - langchain-ai/langgraph-swarm-py

In this blog, I would like to leverage Workato recipes as a tool for the agents to search and book the flights and hotels. The following end-to-end flow diagram illustrates the process.

User Requests Flight

• User asks about flights via Slack.
• Flight Agent receives the request as Flight Agent was set as default active agent.
• Flight Agent searches for flights using Workato’s search_flight recipe.
• Flight Agent shares the results with the user.

User Books Flight

• User asks Flight Agent to book a…