What if you could build an AI engineering team where specialized agents collaborate to solve software development tasks—from planning and implementation to code review and testing?
In this hands-on workshop, participants will build a collaborative multi-agent system using Python and LangGraph. Instead of relying on a single monolithic LLM prompt to handle complex development lifecycles, we will architect a multi-agent state machine where specialized agents take on distinct software engineering roles: a Planner decomposes user stories into actionable specifications, a Developer implements clean code, and a Reviewer verifies the implementation against unit tests and style contracts.
Participants will learn how to orchestrate these agents, manage shared state, integrate tool execution, handle non-deterministic failures, and configure controlled retry loops. Through a practical coding challenge, attendees will explore automated verification loops that turn unstructured LLM generation into reliable, tested software.
Key Takeaways
- Role-Based Agent Design: Architect specialized agents with constrained responsibilities, explicit tools, and isolated context.
- Multi-Agent Orchestration with LangGraph: Manage shared state graphs, conditional branching, and error-recovery transitions.
- Automated Verification Loops: Integrate compiler feedback, automated test runners, and deterministic review criteria.
- Controlled Retries & Guardrails: Prevent runaway execution loops and manage token consumption during autonomous code repair.
Target Audience
- Software engineers, backend developers, technical leads, and engineering managers looking to incorporate collaborative AI agents into development workflows.
Prerequisites
- Working knowledge of Python and familiarity with basic REST APIs or LLM provider endpoints.
