Paperclip

Open-source orchestration layer for running multi-agent “companies” where AIs act as CEO, engineers, designers, and QA — coordinating via a ticketing system, heartbeats, and per-agent budgets. Launched early March 2026; hit 36,000 GitHub stars in three weeks.

Key facts

  • Type: Multi-agent orchestration platform / agent “company” runtime
  • License: MIT (fully open source)
  • Launch date: Early March 2026
  • Popularity: 36,000+ GitHub stars within 3 weeks of launch [src-001]
  • Compatible with: Claude Code, OpenClaw, Codex, Cursor — “bring your own agents”
  • Deployment: Runs locally by default (localhost), VPS-compatible for remote access
  • Founder quote: Co-founder “Dota” cited terminal sprawl (20 concurrent Claude Code sessions) as the core pain point [src-001]

What it does

Paperclip lets you define an org chart of AI agents — a CEO, engineers, marketers, designers, QA — and coordinate them through a web dashboard instead of through individual terminal sessions. Each agent has its own instruction files (agents.md, heartbeat.md, soul.md, tools.md), its own model/provider adapter, and its own monthly budget.

The orchestration model inverts the usual human-in-the-loop pattern. Instead of you chatting with one agent at a time, you act as a board: you set high-level goals, comment on issues, approve or deny hires, and review outputs. The CEO agent delegates actual work — hiring new agents, creating projects, breaking down milestones.

Key features Nate Herk demonstrated [src-001]:

  • Heartbeats — see Heartbeats (Agent Wake-ups). Agents wake up every 4/8/12 hours, re-read their context files, check their task queue, and continue work. Replicates what made OpenClaw proactive.
  • Ticketing system — every conversation, issue, and approval is logged and linkable.
  • Per-agent model choice and budget — CEO can run Sonnet 4.6 while designers run a cheaper model. Hard monthly spend cap per agent.
  • Company templates — pre-built org charts (GStack, Superpowers, Agency Agents) can be imported. One template shipped with 48 agents pre-configured.
  • Skills marketplace — integrates with skills.sh. Paste a GitHub URL or skills.sh path into the company skills panel and all agents get that skill.
  • Routines (beta) — define a scheduled or webhook-triggered workflow assigned to an agent. Cron-like automation over conversation.
  • Multi-company support — run multiple companies in one Paperclip instance (e.g., “Proof Shot” and “AIS Content” side by side).

Related

Source references

  • [src-001] Nate Herk — “Claude Code + Paperclip Just Destroyed OpenClaw” (2026-03-28)

Robin Cartier perspective

This page is part of Robin Cartier's working AI knowledge graph: a practical research layer for production AI, recommendation systems, experimentation, GEO, and agentic web readiness.

The useful next step is to connect this concept back to applied product leadership and operating models.

Recommended next

Keep reading from this thread

From 491 indexed pages and articles.

  1. Wiki concept Agent Orchestration The discipline of coordinating multiple AI agents — each with their own role, context, and tools — so they collectively achieve a higher-level goal Related by paperclip
  2. Wiki concept OpenClaw Peter Steinberger's open-source personal AI agent, initially covered in this wiki as a heartbeat-heavy predecessor to Paperclip but expanded by [src-064] into a Related by paperclip
  3. Insight AI Measurement and Experimentation How to measure AI product impact with evals, adoption metrics, online experiments, guardrails, and cost tracking Readers have engaged with this next