Paperclip Structures AI Agents into Companies

The Breakthrough

Paperclip provides an open-source local control plane that turns individual AI agents into a coordinated 'company' complete with organizational charts, ticket delegation, budgets, heartbeats, and audit logs.

What Actually Worked

• Run npx paperclipai onboard to start local services including Postgres and access the dashboard UI.
• Create a company with a goal such as "build and ship a URL shortener MVP this week", add agents like a CTO and two engineers (one for backend, one for frontend and test coverage), set a token budget, and specify the output code directory path.
• Agents activate on heartbeats; the CTO decomposes the goal into tickets, engineers claim tickets, implement tasks, and commit code while the dashboard tracks delegation ancestry, status changes, spend, and goal alignment.
• Define agent roles and capabilities in SKILLS.md files to prevent vague or rogue behavior.
• Use portable templates, Jira-like dashboard, and self-hosted setup to inspect, modify, and integrate with existing agents.

Context

Individual AI agents excel at isolated code-writing tasks, but multi-agent setups devolve into coordination chaos without ownership, goal memory, or termination signals. Paperclip addresses this by imposing company-like structure around workflows like CrewAI or AutoGen, shifting from prompting single agents to defining company goals, org charts, rules, and budgets. The demo builds a URL shortener MVP, revealing strengths in traceability and cost control alongside issues like slow execution, nonsense tickets from poor rules, persistent token burn, and overkill for simple scripts.

Notable Quotes

• "Raw agents working alone aren't great. Useful but hard to coordinate. Paperclip turns them into a team or I guess in this case it's called a company."
• "Paperclip is the manager, the organizational chart, the ticket board, the budget system, the audit log."
• "If your SKILLS.md files suck your company behaves like a confused startup."
• "This is for building out a lot more, having more of these agents working together. It's definitely worth using but it's not for everything."