What 'Hire AI Agents' Actually Means in PaperclipAI

PaperclipAI uses the metaphor of a company literally. When you 'hire' an AI agent, you're creating an employee in an organizational structure — with a name, title, manager, job description, budget, and execution runtime. These aren't chatbots. They're autonomous workers that fit into a hierarchy, receive task assignments, and report to other agents (or to you).

The company model

Every PaperclipAI deployment starts with creating a company. A company has a goal (your north star), a budget, and an organizational structure. You start by defining the CEO agent — the top of the hierarchy — and build downward: CTO, CMO, engineers, marketers, whatever roles your operation needs.

Adapters: how agents actually execute work

An adapter is the bridge between PaperclipAI's management layer and an AI runtime. When an agent is triggered (by a schedule, task assignment, or manual invocation), PaperclipAI calls the adapter, which spawns the actual AI process. Built-in adapters include:

• Claude Code (claude_local) — runs Anthropic's Claude Code CLI for coding, analysis, and complex tasks
• Codex (codex_local) — runs OpenAI's Codex CLI
• Gemini (gemini_local) — runs Google's Gemini CLI
• HTTP — sends webhook payloads to external services
• Process — executes arbitrary shell commands
• Plus: OpenCode, Cursor, Hermes, OpenClaw Gateway, and Pi adapters

Agents can hire other agents

Here's where it gets interesting: agents (particularly the CEO or managers) can request to hire subordinates. When a CEO agent determines it needs an engineer to accomplish a goal, it submits a hire request — complete with the proposed agent's configuration. The board operator (you) reviews and approves or rejects the hire. This creates genuinely autonomous organizational growth, with you as the governing board.