What it does

Paperclip is a Node.js server with a React UI that orchestrates multiple AI agents—Claude Code, Codex, Cursor, OpenClaw, bash scripts, HTTP bots—under one roof. You define business goals, assign agents to roles in an org chart, set budgets, and let them work on scheduled heartbeats. It looks like a task manager; underneath it’s identity management, ticketing, cost control, and governance.

The interesting bit

The framing is deliberately corporate: agents have job titles, report to managers, and work within token budgets. The README’s analogy is apt—“If OpenClaw is an employee, Paperclip is the company.” The actual cleverness is in the orchestration mechanics: atomic task checkout prevents double-work, persistent state lets agents resume context across heartbeats, and goal ancestry means agents carry the full “why” up the chain.

Key highlights

• Bring-your-own-agent: Works with anything that can receive a heartbeat—Claude, Codex, Cursor, bash, HTTP endpoints.
• Cost throttling: Monthly budgets per agent; spend hits the limit, the agent stops.
• Multi-tenant: One deployment can run multiple isolated “companies” with separate data and audit trails.
• Governance with rollback: Approval gates, versioned config changes, and the ability to pause or terminate any agent.
• Mobile dashboard: React UI works on phone for monitoring autonomous operations.

Caveats

• The README is long on vision and short on concrete setup friction—actual deployment complexity for a “full control plane” with identity, secrets, and multi-company isolation is unclear.
• Star count (68,836) is suspiciously high for a project with no topics and a relatively new niche; no verification of organic growth vs. other factors is provided in the sources.

Verdict

Worth a look if you’re already running multiple AI agents and losing track of them across terminal tabs. Skip it if you’re looking for a single-agent coding assistant—this is coordination infrastructure, not a better autocomplete.