In this project, I build a multi-agent “law firm” that behaves like an AI-powered legal department—end to end. The core idea is simple: instead of one chatbot trying to do everything, I split the workflow into specialized agents that collaborate like real legal roles. One agent focuses on contract analysis, another hunts for legal/compliance gaps, and another summarizes risks into clear, actionable outputs. I run the whole thing with an agentic system-design mindset so you can reuse the same blueprint for other domains too. I also show how I’m thinking about agentic design patterns for legal work: task decomposition, structured outputs, and a clean handoff between agents so the final response doesn’t feel like random LLM text. The demo is about spotting contract risks with high accuracy, generating a legal strategy-style summary, and producing a compliance/risk assessment you can actually use. If you’re new to agents, I point you to my prerequisites first—because without that foundation, you’ll just glue prompts together and call it “agentic.”