I just finished setting up Paperclip AI — an open-source multi-agent orchestration platform — on my Debian machine in Honolulu.
Paperclip lets me run a virtual “company” of AI agents that can collaborate autonomously on tasks. Think of it as building my own JARVIS system, where a CEO agent coordinates specialized agents for research, coding, content creation, cybersecurity tasks, and AWS/cloud work.
What I Built
- Deployed Paperclip as a systemd service so it runs 24/7 in the background, even when I reboot or shut down my laptop.
- Created my first company: NY2HONOLULU AI
- Set up the first agent: NY2HONOLULU CEO (powered by Claude)
- Goal: Combine JARVIS-style intelligence with my real-world focus on AWS development, CSI Linux, cybersecurity, digital forensics, and learning in public.
Why I Chose Paperclip
- Fully self-hosted and open-source (no vendor lock-in)
- Strong governance and budgeting tools to prevent agents from running wild
- Perfect for someone who wants to experiment with autonomous AI teams while keeping everything on their own hardware
The setup had a few hurdles (mostly around Claude Code CLI path and environment variables), but it’s now running reliably as a background service.
Next steps on the blog:
- Full write-up of the installation process
- Custom CEO system prompt I’m using
- First real tasks I’ll give the agent team
- How it integrates with my AWS and CSI Linux workflow
This is another step toward my vision of building practical, locally-controlled intelligent systems from paradise.