AgentMax

AgentMax automates M&A due diligence using six AI agents that collaborate through Band. You submit a company name, and the agents go to work: one searches the web, one reads your uploaded documents, one runs the financials, one checks legal risk, and a final synthesis agent pulls it all together into a scored investment memo with an INVEST, CONDITIONAL, or PASS verdict. Band is the actual coordination layer here. Every handoff between agents happens through Band messages and mentions. The agents pass structured data to each other, escalate to humans when they hit legal red flags, and post their results back to a shared room in real time. It is not a demo wrapper around a single LLM call. On the frontend, there is a Deals Dashboard where you can track every company you have analyzed, filter by recommendation, sort by score, and compare two companies side by side on a radar chart. Analyses persist across sessions so it works like a real portfolio tool. Built with FastAPI, LangGraph, GPT-4o, and the Band SDK. Deployed on Railway. The whole pipeline runs end to end without human input unless the legal agent flags something serious.