hkuds/openopc

OpenOPC is a Python project that lets you build a small company made entirely of AI agents, each one filling a specific job role, working together toward a goal you set. The idea is that instead of hiring people, the system automatically assembles a team of AI employees, assigns them tasks, and coordinates their work, learning from each completed project to improve future ones. The system works in three stages. First, given a goal, it drafts an organization chart, figuring out what roles are needed and how they report to each other, then a recruiter agent either reuses an AI employee from a past project or brings in a new one from a pool of available roles. Second, once the team exists, a manager style process breaks the work into individual items, assigns them, and reviews the results, deciding whether to accept, redo, or escalate each piece. Work that does not depend on other tasks can run at the same time, while dependent work waits its turn. If something blocks progress, the system first tries to resolve it inside the team, and if it is beyond the team's authority, it is passed up to the human who owns the project for a decision. Third, after a run finishes, the system attributes outcomes to the specific roles responsible for them, rather than crediting or blaming the whole company at once, and turns what it learned into longer term organizational knowledge. OpenOPC covers a wide range of use cases across nine areas, including AI development, software engineering, financial investment research, sales, video and content production, e-commerce, accounting, education, and various industry specific assistants. It includes both a command line interface and a browser based office style interface built with React and a game engine called Phaser, where you can watch the AI team's kanban board and activity in something like a virtual office. The project is released under the MIT license. The full README is longer than what was shown.