Skip to main content
PromptQuorumBuilt for humans. Structured for AI.
Home/Power Local LLM
Local AI Agents & Tool Use

Paperclip Review: Open-Source Control Plane for AI Agent Teams

·8 min readBy · Founder of PromptQuorum · Discovery engine for open-weight & open-source AI

Paperclip is a free, MIT-licensed, self-hosted control plane that manages teams of AI agents you bring yourself, installed with one npx command from paperclip.ing. It is a manager, not a model, so what it costs and what leaves your machine depends on the agents you connect.

Key Takeaways

  • It runs agents you bring: the README lists OpenClaw, Claude Code, Codex, Cursor, Gemini CLI, OpenCode, Pi, Hermes, Grok Build and Kimi Code, and ships no model
  • Control features: per-agent budgets with warning thresholds and hard stops, approval stages for hires and task reviews, and an activity log
  • Install is one command, npx paperclipai@latest onboard --yes, with an embedded PostgreSQL database and data kept locally
  • The terms say self-hosted installs send anonymized telemetry, and richer telemetry is opt-in
  • The README roadmap lists a desktop app as not started; today the interface is a web app

📍 In One Sentence

Paperclip is a free, MIT-licensed, self-hosted server that gives AI agents an org chart, goals, budgets and approval gates, so you manage the work instead of many separate terminals.

💬 In Plain Terms

You describe a goal, add agents such as Claude Code or Codex as employees, and Paperclip wakes them with tasks, tracks what they spend and asks you to approve key steps.

Paperclip (paperclip.ing, source code at github.com/paperclipai/paperclip) is a free, open-source Node.js server with a React interface that organizes AI agents into a company: an org chart, goals, tasks, budgets and approval steps. It does not include a model; you connect agents you already use. This review covers what it does, how to install it, its license, telemetry and privacy statements, and how it differs from other agent tools in the directory. PromptQuorum has not tested Paperclip hands-on; every statement comes from the maker's pages and the GitHub repository, checked on 19 November 2026.

Paperclip Review: Open-Source Control Plane for AI Agent Teams

What Is Paperclip?

The GitHub README calls Paperclip "the app people use to manage AI agents for work" and "open-source orchestration for teams of AI agents." It describes a Node.js server and React UI that look like a task manager and add org charts, budgets, governance and goal alignment underneath.

  • Maker: the website footer and the terms of service name Paperclip Labs, Inc.; the project is MIT-licensed on GitHub
  • Documentation: guides, CLI reference and API reference live at docs.paperclip.ing
  • Current release: v2026.1117.0, whose notes describe a new default workspace with Agent Chat, task chat that shows an agent's plan, and runs that survive restarts
  • What it is not: the README says it is not an agent framework, not a chatbot and not a prompt manager; agents bring their own prompts, models and runtimes

What Paperclip Does

Per the README, the feature set covers the organization, the work, the cost and the connections around a team of agents.

  • Org chart and heartbeats — agents get roles, titles and reporting lines, and wake for assigned work, follow-up messages or schedules; delegation flows up and down the chart
  • Tasks and goals — tickets carry company, project and goal links, comments, documents and run history, so agents see the reason behind a task
  • Budgets — spend tracking by company, agent, project and model, with warning thresholds and hard stops; the README notes that enforcement uses recorded spend, so reporting delays and in-flight work can postpone a stop
  • Governance — approval workflows for hires, review stages for task completion, and pause, resume or terminate controls for agents
  • Routines — recurring work on a schedule or triggered by an API call or webhook, each run creating a tracked task
  • Connections and skills — services such as GitHub, Notion and Railway or your own MCP server, with actions set to Allowed, Ask first or Off, plus shared skills in a Skill Studio
  • Multi-organization and portability — several organizations in one deployment, and export or import of agents, skills and optional projects with secret values omitted

Get Paperclip

There is no installer or desktop app: the easiest path is one terminal command, npx paperclipai@latest onboard --yes, with Node.js 24.11 or newer installed. A terminal and a developer-style setup are required.

Option
How to get it
npx (easiest)npx paperclipai@latest onboard --yes
Managed installnpx paperclipai install, adds a paperclipai command
From sourcegit clone, then pnpm install and pnpm dev
Websitepaperclip.ing
Privacy policyPrivacy policy
Documentationdocs.paperclip.ing
Source code (GitHub)paperclipai/paperclip

The current version is 2026.1117.0, published on 18 November 2026, and the last upstream update was the latest commit on 19 November 2026, per GitHub, checked 19 November 2026. The release lists no installer files, and the README roadmap lists a desktop app as not started. The pages read do not state supported operating systems.

Install with npx

These steps follow the README quickstart and the installation guide, checked 19 November 2026; PromptQuorum has not run them.

  1. 1
    Check Node.js
    Why it matters: The guide requires Node.js 24.11.0 or newer on your PATH; older versions get a warning or, for a managed install, a refusal.
  2. 2
    Run onboarding
    Why it matters: Run `npx paperclipai@latest onboard --yes`. The CLI runs from npm's cache and your instance data stays in `~/.paperclip` by default, using an embedded PostgreSQL database unless you supply your own.
  3. 3
    Decide on a background service
    Why it matters: Onboarding can install Paperclip as a service that starts on login; the flags `--install-service` and `--no-install-service` skip the prompt.
  4. 4
    Bring your agents
    Why it matters: The website says local Claude or Codex subscription sign-in needs Python 3 and the provider CLI; for a throwaway instance the README offers `npx paperclipai test-drive` with an API key.

Pricing, License, and Privacy

The README says Paperclip is open source, self-hosted and needs no Paperclip account; the agents and model providers you connect bill you separately.

  • License: MIT for the open-source project; the terms say the hosted services of Paperclip Labs, Inc. are separate and governed by those terms
  • Hosted version: the website and README offer a Paperclip Cloud waitlist and show no price
  • Operator and law: the terms are governed by Delaware law with arbitration under American Arbitration Association rules, and no postal address or imprint was found
  • Telemetry: the terms say self-hosted installs collect anonymized usage, performance and error data, and that opt-in detailed telemetry (stack traces, run logs, configuration) may be used for any purpose related to operating and improving the services, including training machine learning models
  • Hosted-service data: the privacy policy says Paperclip Cloud collects account identifiers, usage data and session recordings, and that data may be processed in the United States
  • Connected accounts: the policy says Google user data is shared with AI providers only to perform tasks you give your agents

Who Should Use Paperclip?

Paperclip fits people who already use several agents and are comfortable running a Node.js service.

Use Paperclip if

  • You run many agent sessions and want tasks, budgets and approvals in one place
  • You want an open-source manager you host yourself
  • You are comfortable with a terminal and Node.js

Look elsewhere if

  • You want a desktop installer or a built-in model
  • You need a vendor with a published address as data processor
  • You do not want any telemetry from a self-hosted install

Paperclip vs. Other Agent Tools

Paperclip manages agents rather than being one, so the closest comparisons are other agent orchestration and workbench tools. See the Local AI Directory for the full catalog.

  • OpenClaw — an agent Paperclip can employ; the README says "If OpenClaw is an employee, Paperclip is the company." See the OpenClaw review.
  • MyAgents — a free desktop workbench that runs agent runtimes and schedules recurring tasks, with installers for macOS and Windows. See the MyAgents review.
  • OpenWork — a free desktop app in which AI agents work on your own files, built on OpenCode. See the OpenWork review.
  • CrewAI — a code framework for role-based multi-agent workflows rather than a ready-made dashboard. See the CrewAI review.

Compare in directory

Frequently Asked Questions

Does Paperclip include an AI model?

No. Each agent brings its own model and runtime, so costs and data flow follow the agents you connect; the README describes Paperclip as the organization around them.

Can I try it without touching my real setup?

The README describes a test-drive command that starts an isolated, already-initialized instance in a temporary data directory and needs a provider API key.

Does it run on Windows, macOS and Linux?

The pages read do not list supported operating systems; they require Node.js 24.11 or newer and describe a web interface served from your own machine.

Can I manage it from a phone?

The README lists "Mobile Ready" and the website mentions a pnpm dev:mobile script, but no separate mobile app was found.

Sources

← Back to Power Local LLM

MCP