Descriptions:
Paperclip is an open-source AI agent orchestration platform that reached 38,000 GitHub stars in under three weeks, and this tutorial covers a full setup from scratch. Unlike single-agent tools, Paperclip sits above individual agents as an orchestrator: it lets you build a team with defined roles (CEO, researcher, writer, editor), individual token budgets, and a shared task system modeled after project management software — every action and every token spent is logged in a central dashboard.
Youri van Hofwegen walks through deploying Paperclip on a VPS (recommended: KVM2 with 2 CPU cores and 8GB RAM) so agents run continuously without depending on a local machine. Setup involves connecting API keys for Anthropic, OpenAI, Gemini, or Cursor, with support for mixing models — for example, Claude for the reasoning-heavy CEO agent and a cheaper model for formatting tasks. Paperclip supports Claude Code and Codex adapters, so existing subscriptions plug in directly without additional cost.
A live demo shows a three-agent content operation: a research agent that browses for trending AI topics and compiles structured briefs with talking points and sources, a writer agent that turns briefs into full articles, and an editor agent that reviews before publication. By default, the CEO agent requires human approval before hiring new agents — a control gate that can be disabled as confidence builds. The video covers the heartbeat checklist system each agent follows on wake cycles and how to write effective per-agent instruction sets for consistent output.
📺 Source: Youri van Hofwegen · Published April 17, 2026
🏷️ Format: Tutorial Demo
