Key takeaways

  • Run many AI agents from one dashboard, with goals, tasks and an org chart, instead of juggling terminal tabs.
  • Cap what agents cost: set budgets that track reported spend and can pause work at your limits.
  • Keep a person in charge: approval gates stop agent work going ahead unchecked, and config changes can be rolled back.
  • Bring the agents you already have. Any agent that can receive a heartbeat can join the team.
  • The catches: telemetry is on by default, budget stops can lag, and some features are still experimental.

What it is

Paperclip describes itself as open-source orchestration for teams of AI agents. It is a self-hosted Node.js server and React interface, released under the MIT licence, and the code is on GitHub at paperclipai/paperclip.

The problem it solves

Anyone who runs several coding agents knows the mess. The project's own pitch is that you have 20 Claude Code tabs open and cannot track which one does what.

Paperclip is not another way to build agents. It does not tell you how to build them; it manages the organisation they work in, so work and costs sit in one place.

What you get

  • Use the agents you already have: anything that can receive a heartbeat can be added to the team.
  • Fewer clashes: a single assignee and execution locks stop competing runs claiming the same task.
  • Spend you can see: budgets track reported spend and can pause work at configured limits.
  • Oversight: approval gates are enforced, and configuration changes are revisioned and can be rolled back.
  • Room to grow: one deployment can host many organisations with separate data.

The catches

  • Privacy: anonymous usage telemetry is collected by default, though the project documents a setting to turn it off.
  • Budgets are not instant: enforcement uses recorded spend, so reporting delays and work already in flight can delay a stop.
  • Maturity: Agent Chat and the chat connectors for Slack, Discord, Telegram and AgentMail are experimental, and bringing your own ticket system is still on the roadmap.
  • Requirements: it needs Node.js 24.11 or newer, and sandbox support depends on the environment you configure.

What we did not test

We did not run Paperclip, connect any agent or check its budget limits. Every claim here comes from a literal quote in the project's README or the GitHub API, and the feature descriptions are the project's own.

Verdict

Editors' view: this is aimed at people who already run several coding agents and are losing track of the work and the spend. If you use a single agent, it is probably more structure than you need. If you try it, do so in a throwaway instance with low budgets.

Repository
paperclipai/paperclip
TypeScript · MIT
View on GitHub ↗

Sources

Need help adopting this, or have a software project in mind?

Tell us what you are building. The initial consultation is free. Email [email protected] or follow @zolnaltd.

Start a conversation →