What it is
Paperclip is a self-hosted Node.js server and React UI that orchestrates a team of AI agents to run a business. You bring your own agents (e.g. OpenClaw, Claude Code, Codex, Cursor, Gemini CLI), assign goals, and track work and costs from one dashboard. It looks like a task manager but adds org charts, budgets, governance, goal alignment and agent coordination.
Who it's for
- People coordinating many different agents toward a common goal who lose track of many simultaneous terminals
- Teams that want agents running autonomously 24/7 while keeping audit and approval control
- Users who need to monitor costs and enforce budgets across agents
- Operators who want to manage autonomous businesses from a phone
Requirements
Requirements
- Node.js 24.11 or newer
- Provider API key (e.g. ANTHROPIC_API_KEY, OPENAI_API_KEY, OPENROUTER_API_KEY) for the test-drive examples
Setup
Install and onboard
With Node.js 24.11 or newer installed, run the onboarding CLI. It runs from npm's cache; instance configuration and data persist locally.
bashnpx paperclipai@latest onboard --yesTry an isolated test instance
test-drive creates an isolated manual test instance already initialized with a CEO agent. It stays in the foreground, never installs a service or creates a first task, and opens the browser only after setup succeeds.
bashANTHROPIC_API_KEY=... npx paperclipai test-drive
Examples
Test drive with Codex harness
bashOPENAI_API_KEY=... npx paperclipai test-drive --harness codexWhat it does: Starts an isolated test instance using the codex harness with an OpenAI key.
Test drive with OpenCode and OpenRouter
bashOPENROUTER_API_KEY=... npx paperclipai test-drive \
--harness opencode \
--model openrouter/anthropic/claude-sonnet-4.5What it does: Selects the opencode harness and a specific model routed through OpenRouter.
Define a company goal
PromptBuild the #1 AI note-taking app to $1M MRR.Expected output: The README's example of step 01 (define the goal), followed by hiring a team and approving/running from the dashboard.
Pros & cons
Pros
- Pro:Supports many agent runtimes (OpenClaw, Claude Code, Codex, Cursor, Gemini CLI, OpenCode, etc.) plus custom processes, HTTP endpoints and external adapters
- Pro:Built-in budgets with threshold alerts and hard stops, plus governance, approval gates and revisioned config with rollback
- Pro:Atomic task checkout and persistent task context, comments and documents prevent duplicate work and lost state
- Pro:Self-hosted, open source (MIT badge), no Paperclip account required
Cons
- Con:Budget enforcement uses recorded spend, so usage reporting and in-flight work can delay a stop
- Con:Agent Chat and chat/email connectors are experimental and require enabling instance settings
- Con:Exported company packages need review before sharing since plain environment values and local paths can remain
Images
