Paperclip
A control room for your AI agents: org chart, budgets, and approvals in one dashboard
111k+
GitHub stars
10
agent tools it can run, including Claude Code, Codex, and Cursor
2,700+
pull requests merged, according to the maintainers
What it actually does
If you run several AI agents at once, it gets hard to say who is doing what, what it costs, and whether the work is any good. Paperclip is a free dashboard that sits above them. You set a goal, add agents to an org chart, and follow their work, spending, and approvals from one page in your browser. The maintainers call it managing business goals instead of pull requests. It is MIT licensed and runs on your own machine.
Why it's worth your time
- Puts every agent you run on one dashboard, with an org chart, task threads, and a log of what each one did.
- Works with the agents you already use: Claude Code, Codex, Cursor, Gemini CLI, OpenCode, and more.
- Budgets can be set per company, per agent, or per project. You get a warning at a threshold and work pauses at the limit.
- Approval stages let you sign off on plans and new hires before an agent starts, and you can pause, reassign, or stop any of them.
- Recurring routines run on a schedule or a webhook, and each run becomes a tracked task with an owner and a history.
- Starts with one command. Your data stays on your own computer, and a local database is set up for you.
Before you start
- Node.js 24.11 or newer
- At least one agent you already use, such as Claude Code or Codex, for Paperclip to run
- Python 3 and that agent’s command-line tool installed, only if you want to sign in with a Claude or Codex subscription from the same machine
🆕 New to the command line? Start here
Paperclip isn't a Claude feature -- it runs completely on its own, so you don't need Claude Code or any AI subscription to use it. All you need is a terminal, a plain text window that's already built into your computer.
1. Open a terminal
- Mac: press
Cmd + Space, type "Terminal", press Enter. - Windows: press the Windows key, type "Terminal" or "PowerShell", press Enter.
2. Install Node.js 24.11 or newer first -- the runtime the npx quick-start command below uses
Mac
brew install nodeWindows
winget install OpenJS.NodeJS3. Then paste the install command below into that same window and press Enter. That's the whole install.
Already use Claude Code (the desktop app, claude.ai/code in a browser, or the CLI)? You can skip opening a terminal yourself -- paste the install command below into a Claude Code chat and ask it to run it for you, it has real terminal access built in. Claude Cowork specifically can't do this on its own -- it's built for working with files and documents, not for running command-line tools.
Install it
Quickest -- one command
npx paperclipai@latest onboard --yesReach it from other devices on your Wi-Fi (asks you to sign in)
npx paperclipai@latest onboard --yes --bind lanFrom source
git clone https://github.com/paperclipai/paperclip.git
cd paperclip
pnpm install
pnpm devUse it
Open the dashboard once it is running
http://localhost:3100Try it with a throwaway sample company first
ANTHROPIC_API_KEY=... npx paperclipai test-driveTurn off anonymous usage data
PAPERCLIP_TELEMETRY_DISABLED=1 npx paperclipai@latest onboard --yesFrequently asked questions
Is Paperclip free?
Yes. Paperclip is MIT licensed and free to install and run. The agents you connect have their own costs, such as a Claude or Codex subscription or API usage, and Paperclip is the place where you track and cap that spending.
Does it come with its own AI agents?
No. You bring the agents. The project’s rule of thumb is that anything that can receive a heartbeat can be hired, which covers Claude Code, Codex, Cursor, Gemini CLI, OpenCode, custom scripts, and web endpoints.
Where does my data go?
Settings and data are saved on your computer, and a PostgreSQL database is set up for you. Anonymous usage data is collected by default, and the README says it contains no personal information, prompts, or secrets. You can turn it off with PAPERCLIP_TELEMETRY_DISABLED=1.
Will a budget limit stop an agent instantly?
Not always. The README says enforcement depends on recorded spending, so delayed usage reports and work already in progress can hold up a stop. Set limits with some headroom.
Is it safe to open to the internet?
The quick start runs in a trusted local mode meant for your own machine. For remote or shared access, the README points to authenticated mode on a private network such as Tailscale.
Do I need to be a developer to use it?
You need to run one command in a terminal and have Node.js installed, and the agents it manages need their own setup. After that, day-to-day use happens in a browser dashboard.