From Zero to CEO: Building Your First Automated Dev Shop in 10 Minutes
A practical, step-by-step tutorial to get users from installation to their first running 'company'.
Vijayaragupathy
AI Engineer, ML systems builder, and applied agentic workflow developer

From Zero to CEO: Building Your First Automated Dev Shop in 10 Minutes
Paperclip is designed to be approachable. You can go from zero to running your first "AI company" in under 10 minutes. Here's how.
Prerequisites
Before we begin, you'll need:
- Node.js 18+: Install from nodejs.org
- API Keys: OpenAI, Anthropic, or other provider keys (Paperclip supports multiple providers)
- Terminal access: To run commands
Step 1: Installation
npx paperclipai onboardThis command will:
- Check your environment
- Prompt for your API keys
- Create a new Paperclip project
- Open the dashboard
The onboarder will guide you through the setup process with clear prompts.
Step 2: Defining Your Identity
Every Paperclip organization starts with an identity. Create a file named identity.md in your project root:
# Company Identity
- **Name**: My AI Company
- **Mission**: Build amazing products with AI
- **Values**: Speed, quality, reliability
- **Constraints**: Budget $50/day, no external deploymentsThis file defines the "what" and "why" of your AI company.
Step 3: Defining Your Agents
Create agents.md to define your workforce:
# Agent Roles
## CEO Agent
- **Role**: High-level strategy and decision making
- **Tools**: read, write, search
- **Budget**: $10/day
- **Heartbeat**: Every hour
## Engineer Agent
- **Role**: Implementation and coding
- **Tools**: exec, read, write, git
- **Budget**: $20/day
- **Heartbeat**: Every hour
## Auditor Agent
- **Role**: Quality control and cost monitoring
- **Tools**: read, exec
- **Budget**: $5/day
- **Heartbeat**: Every 6 hoursEach agent has:
- A clear role and purpose
- Allowed tools (prevents unauthorized actions)
- A budget cap (prevents runaway costs)
- A heartbeat schedule (when they run autonomously)
Step 4: Setting Up the First Heartbeat
Paperclip's "heartbeat" is a scheduled execution cycle. Create a heartbeat definition:
name: "daily-build"
schedule: "0 9 * * *" # Daily at 9 AM
agents:
- CEO
- Engineer
- Auditor
goal: "Review pending issues and prioritize next development tasks"This heartbeat will run every day at 9 AM, coordinating your agents to work together.
Step 5: Launching the Dashboard
Once configured, start the Paperclip dashboard:
paperclipai dashboardThis will open a local web interface where you can:
- Monitor agent activities
- View logs and events
- Adjust budgets and schedules
- Set new objectives
Step 6: The First Goal
Now, give your CEO agent its first objective:
Current Objective:
- Review the backlog of 5 pending issues
- Prioritize based on business value and technical complexity
- Create a development plan for the next sprintWhen the heartbeat runs, your CEO will:
- Read the backlog
- Analyze and prioritize
- Create a development plan
- Hand off tasks to the Engineer agent
What's Next?
You now have a running "AI company" that operates autonomously:
- Heartbeats run on schedule
- Agents work within their defined roles
- Governance ensures safety and cost control
- Dashboard gives you visibility into everything
Next Steps
- Add more agents: Create specialized roles for different tasks
- Customize heartbeats: Create recurring workflows
- Integrate with external tools: Connect to GitHub, CI/CD, etc.
- Scale up: Start with a single agent, grow to a full organization
Common Patterns
Pattern 1: The "Build and Review" Cycle
name: "build-review"
schedule: "0 */6 * * *" # Every 6 hours
agents:
- Engineer
- Auditor
goal: "Build requested features and review code quality"Pattern 2: The "Research and Report" Cycle
name: "research"
schedule: "0 10,16 * * *" # 10 AM and 4 PM
agents:
- Researcher
goal: "Research industry trends and prepare weekly report"Pattern 3: The "Monitor and Alert" Cycle
name: "monitor"
schedule: "*/30 * * * *" # Every 30 minutes
agents:
- Monitor
goal: "Monitor system health and alert on issues"Troubleshooting
Dashboard won't start
Make sure port 3000 is available:
lsof -i :3000Agent not running
Check the dashboard logs for errors. Common issues:
- API keys not configured
- Budget exceeded
- Tools not allowed for agent role
High costs
- Reduce agent budgets
- Tighten tool permissions
- Increase heartbeat frequency (reduce scope per run)
Conclusion
With just a few files and a dashboard, you've built an autonomous AI organization. The key is to start simple, iterate, and gradually increase complexity.
Paperclip handles the heavy lifting—governance, scheduling, observability—so you can focus on defining great objectives and watching your AI company grow.
Next: The ROI of Autonomy
Continue Reading
More from the system
Orchestration
Beyond the Chatbot: Orchestrating Entire AI Companies with PaperclipThe paradigm shift from single-agent task completion to multi-agent organizational management.
Autonomy & Strategy
The ROI of Autonomy: Why Paperclip is the Infrastructure for the 1-Person UnicornThe tangible business and operational advantages of using Paperclip over manual agent management.
Benchmarks & Evaluations
Orchestration vs. Automation: How Paperclip Differs from CrewAI, AutoGPT, and LangGraphPositioning Paperclip in the current AI ecosystem and clarifying its unique value proposition.