Paperclip organizes AI agents into org charts that mirror real companies. Every agent has a role (CEO, manager, or individual contributor), a single reporting line, and delegated authority over the agents beneath it. The org chart isn't just for show -- it controls how work flows through the system, who can assign tasks to whom, and how decisions escalate when an agent gets stuck.
Without this structure, you end up with a flat pool of agents tripping over each other. No one knows who assigned what, duplicate work piles up, and there's no clear escalation path when something breaks. The org chart fixes that. The CEO sets strategy, managers coordinate their teams, ICs execute tasks -- and Paperclip tracks all of it.
Master paperclip org charts & agent hierarchy
Take the Paperclip Fundamentals course with hands-on lessons and challenges.
The org chart API returns a nested tree starting from the CEO. Each node has the agent's ID, role, adapter type, and its direct reports. This is the structure that drives delegation and work assignment.
The reportsTo field sets the reporting line. Note that agent creation goes through governance -- the board has to approve the hire before the agent goes live. Each agent also gets its own monthly token budget.
When an IC checks out an issue, it doesn't just get the task description. It gets the full chain: which project, which goal, which company objective. The agent knows why the work matters, not just what to do. That context flows down through the org chart automatically.
Managers do strategic thinking -- breaking goals into projects, assigning issues, reviewing output -- so they benefit from heavier models. ICs do focused execution, so lighter models usually work fine. The adapter system lets you mix and match runtimes across the org chart.
Creating a flat org chart where every agent reports directly to the CEO, overloading the CEO with coordination work
Add manager agents between the CEO and ICs. The CEO should delegate to 2-4 managers, each running their own team. Same way real companies scale -- you don't have the CEO micromanaging every engineer.
Assigning the same adapter and model to every agent regardless of role, wasting budget on overpowered models for simple tasks
Match model power to role complexity. CEO and managers benefit from powerful reasoning models (Claude Opus, GPT-4o). ICs doing focused tasks can use faster, cheaper models (Codex Mini, GPT-4o-mini).
Trying to create multiple CEO agents in a single company, which violates the single-CEO constraint
Each company has exactly one CEO. If you need parallel leadership, create separate companies within the same Paperclip instance -- each gets its own CEO and org chart.
Bypassing the governance system by directly modifying agent reporting lines through the database instead of the API
Always use the API for structural changes. Governance approval exists to prevent autonomous runaway -- directly editing the database circumvents these safety checks.
The org chart is how Paperclip turns a pile of independent agents into a coordinated team. CEO at the top, managers in the middle, ICs doing the work -- each with a single reporting line and no ambiguity about who does what. The key things to get right: plan the hierarchy before creating agents, use heavier models for strategic roles and lighter ones for focused tasks, and don't skip governance even when it feels like overhead.
Interactive lessons and challenges on Stanza, practice in VS Code, Cursor, or the web.
Interactive lessons and challenges, right in your code editor.
Check the free courses. No credit card.