Paperclip gives your agents an org chart and a budget
Trending hard on GitHub today at +743, sitting on 77.1k stars and 14.2k forks. Paperclip calls itself the app people use to manage AI agents for work, and the framing is deliberately boring: it is a control plane, not a chat window. You define what the business is trying to do, assign agents to roles, and watch the work and the money from one dashboard.
The features read like an HR and finance system, which is the point. Task management. Org charts. Budget enforcement. Governance controls. Execution monitoring. It drives Claude Code, Codex, and whatever custom agents you bring, so it is not competing with your harness β it sits above it and decides which harness gets which job and how much it is allowed to spend doing it.
MIT licensed, TypeScript and Node with a React frontend, 3,545 commits on master. Also 2,100 open issues, which tells you both that a lot of people are using it and that this is a young project moving fast. Do not deploy it Friday afternoon.
This is the third entry in a pattern that started with T3 Code and OpenChamber last week. Nobody is trying to build a better agent right now; everyone is trying to build the layer that tells a fleet of agents what to do and stops them from burning ten thousand dollars overnight. Budget enforcement as a headline feature is the tell. A year ago the interesting question was whether the agent could finish the task. Now the interesting question is what it cost and who approved it.
https://github.com/paperclipai/paperclip
← Back to all articles
The features read like an HR and finance system, which is the point. Task management. Org charts. Budget enforcement. Governance controls. Execution monitoring. It drives Claude Code, Codex, and whatever custom agents you bring, so it is not competing with your harness β it sits above it and decides which harness gets which job and how much it is allowed to spend doing it.
MIT licensed, TypeScript and Node with a React frontend, 3,545 commits on master. Also 2,100 open issues, which tells you both that a lot of people are using it and that this is a young project moving fast. Do not deploy it Friday afternoon.
This is the third entry in a pattern that started with T3 Code and OpenChamber last week. Nobody is trying to build a better agent right now; everyone is trying to build the layer that tells a fleet of agents what to do and stops them from burning ten thousand dollars overnight. Budget enforcement as a headline feature is the tell. A year ago the interesting question was whether the agent could finish the task. Now the interesting question is what it cost and who approved it.
https://github.com/paperclipai/paperclip
Comments