Manage AI agent teams like a company with this open-source dashboard
When multiple AI agents work simultaneously, it is difficult to see who is doing what, how much it costs, and whether approval procedures are being followed. Paperclip is a tool that places these agents in an org chart and assigns them goals and budgets for management. It provides an environment where you can direct and audit an agent team, much like the executives of a real company.
Four pillars for agent organizations
Paperclip is designed around four elements necessary for an AI agent organization to actually deliver results. First, a task management system where agents declare intent, perform tasks, and verify output. Second, an org chart that assigns roles, permissions, and boundaries to both humans and agents. Third, a training system that designs and evaluates agent skills and reviews performance. Fourth, infrastructure that supports various models and runtimes while handling security and cost control. These four components must work together for an agent team to operate stably.
Control tasks, costs, and permissions in one place
The core of this tool is controlling distributed agent activity from a single dashboard. Each agent has a boss and rank in the org chart, and tasks are issued as tickets with attached conversations. Budgets are set at the company, agent, and project levels, and work is automatically paused when spending limits are reached. You can also set approval stages and governance rules to enforce human intervention for important decisions. Through this structure, agents can work autonomously while accountability remains clear.
Installation and execution environment
Paperclip requires Node.js 24.11 or higher and pnpm 9.15 or higher. You can start quickly with the npx paperclipai@latest onboard --yes command, which automatically creates an embedded PostgreSQL database. When running locally, a trusted loopback mode is applied by default, while remote access can use private network binding like Tailscale or Docker deployment. You must set API keys for services such as Anthropic and OpenAI to communicate with agents.
Situations where this tool is suitable
This tool is useful when operating multiple AI agents simultaneously makes task tracking difficult, or when you need to control and audit agent spending. It is particularly suitable for teams building agent teams that perform various roles such as marketing, design, and operations, in addition to coding agents. However, if you only want to use a single agent simply or focus on prompt engineering for the agent itself, the organizational management features of this tool may be excessive. It also delivers the greatest effect when you want an operational style where agents work autonomously 24/7 but you only intervene when necessary.
