What if an entire company were run by AI agents — complete with budgets, hierarchies, governance, and audit trails? That’s exactly what Paperclip AI makes possible.
What is Paperclip AI?
Paperclip is an open-source orchestration platform for so-called “Zero-Human Companies” — organizations run entirely by AI agents. It’s neither a chatbot nor a traditional agent framework. Instead, it’s a management layer that sits on top of existing AI agents, coordinating, monitoring, and budgeting them.
The project is available under the MIT license on GitHub with over 21,000 stars and is actively maintained. The current version is v0.3.1 (March 2026).
The Core Concept: Bring Your Own Agent
Unlike platforms such as CrewAI or AutoGen that come with their own agent definitions, Paperclip follows a “Bring Your Own Agent” approach. This means you can plug in any AI runtime — from Claude Code and OpenClaw to Codex, Cursor, or even simple Bash scripts.
Agents communicate through an intelligent heartbeat system: they wake up, receive tasks, complete them, and report back. Through Runtime Skill Injection, agents can learn new workflows on the fly — without any retraining.
Organizational Structure Like a Real Company
What makes Paperclip unique is how consistently it maps corporate structures onto AI agents:
- Org charts with hierarchies, roles, titles, and reporting lines
- Departments like Engineering, Finance, and Operations
- Projects and issues — every task is traceable back to the company mission
- Multi-company support — a single deployment can run multiple isolated “companies” simultaneously
In practice, this means you can define a CEO agent that makes strategic decisions and delegates tasks to a CTO agent, who in turn coordinates engineering agents. Each agent has a clearly defined area of responsibility.
Budget Control: No Agent Runs Unchecked
A common problem with autonomous AI systems is spiraling costs. Paperclip solves this with built-in budget management:
- Monthly budgets per agent with hard enforcement
- Automatic pause at 100% utilization, warning at 80%
- Cost tracking at task, project, and company level
- Transparent billing: every API call is logged
If you’re running 20 Claude Code sessions in parallel, you’ll always know exactly what each agent costs — and can throttle or pause individual agents as needed.
Governance and Control
Paperclip treats human users as “Board Members” with full control over the AI organization:
- Approval gates for strategic decisions — critical actions require human sign-off
- Approve, pause, terminate, or override agents at any time
- Versioned configurations with rollback capability
- Immutable audit log (append-only): every tool call, every instruction, every decision is recorded
This makes Paperclip particularly attractive for scenarios where traceability and compliance matter.
Technical Architecture
Paperclip is written in TypeScript and uses the following technologies:
- Backend: Node.js 20+ with Express
- Frontend: React-based dashboard
- Database: PostgreSQL (embedded or external)
- API: REST on port 3100
- Deployment: Docker, Docker Compose, or natively via npx
The setup is surprisingly simple:
# Quick start with npx (embedded PostgreSQL)
npx paperclipai onboard --yes
# Or via Docker Compose
docker compose -f docker-compose.quickstart.yml up -dUse Cases
1. Autonomous Software Company
Organize multiple coding agents as a team: a CTO agent plans the architecture, engineering agents implement features, and a test engineer validates the results. All coordinated through Paperclip’s issue system.
2. Multi-Business Management
Anyone running multiple projects or business models in parallel can create a separate “company” in Paperclip for each — with isolated budgets, agents, and governance rules.
3. Smart Home and Infrastructure
Specialized agents monitor and optimize home infrastructure: a Smart Home agent manages Home Assistant automations, while a DevOps agent handles Docker containers and backups.
4. Content and Reporting
Agents that regularly generate reports, create social media posts, or analyze data — with clear budget limits and human approval before publication.
5. 24/7 Operations
Thanks to the heartbeat system, agents work autonomously around the clock. They wake up, check for new tasks, complete them, and go back to standby — without human intervention.
How Does Paperclip Compare to Other Frameworks?
| Aspect | Paperclip | CrewAI / AutoGen |
|---|---|---|
| Focus | Organization & Governance | Workflow Execution |
| Agents | Bring Your Own (any runtime) | Built-in definitions |
| Budget | Built-in, per agent | Not available |
| Multi-Company | Yes, isolated | No |
| Audit | Append-only log | Minimal |
| Metaphor | “Found a company” | “Define a workflow” |
Paperclip doesn’t compete directly with agent frameworks — it sits one layer above. You use Paperclip to coordinate and monitor agents running on any runtime of your choice.
Real-World Example: Paperclip on a Raspberry Pi
Paperclip runs perfectly as a Docker container on a Raspberry Pi 5 — including its PostgreSQL database. In my setup, Paperclip manages eight specialized agents:
- CEO — strategic coordination and prioritization
- CTO — technical architecture decisions
- CFO — budget monitoring and cost optimization
- Python Expert — scripting and automation
- Java Expert — backend development
- Test Engineer — quality assurance and testing
- Smart Home Expert — Home Assistant and IoT integrations
- Financial Expert — analytics and reporting
All agents use Claude Sonnet as their backend and have access to the local network, Docker containers, and SSH connections. The heartbeat system ensures agents only become active when there are actual tasks to complete — saving both resources and costs.
Getting Started
Ready to try Paperclip? Here’s how to get started in under five minutes:
# Option 1: Quick start (includes embedded PostgreSQL)
npx paperclipai onboard --yes
# Option 2: Docker (recommended for servers)
git clone https://github.com/paperclipai/paperclip.git
cd paperclip
docker compose -f docker-compose.quickstart.yml up -d
# Then open http://localhost:3100 in your browserFrom there, you can create your first company, define agents, assign roles, and start delegating tasks.
Conclusion
Paperclip AI fills a gap in the AI ecosystem: while other tools focus on executing individual agents, Paperclip provides the organizational infrastructure for entire AI teams. With budget management, governance, audit logs, and the flexible “Bring Your Own Agent” approach, it’s the ideal platform for anyone looking to move beyond single chatbot interactions.
Whether as an experiment on a Raspberry Pi or as a production system in the cloud — Paperclip makes the leap from “I use AI” to “AI works for me” tangible.
Leave a Reply