How J Donmoyer Runs a Private OpenClaw Assistant on a Raspberry Pi 5
Built by Small Business Owner · · Source: Blog
Result: By the author's estimate, about 3 hours saved on day one, on roughly $100 of hardware and about $30 a month in Venice AI credits.
What they built
J Donmoyer, who writes the Transmutable Explorations blog, set up a self-hosted AI assistant for their small business on a Raspberry Pi 5 with 4GB of RAM. They say the hardware cost about $100 one time. The Pi runs OpenClaw, an open source AI gateway, and sends requests to the Kimi K2.5 model (256K context window) through the Venice AI API. Tailscale handles remote access. The goal, according to the author, was an assistant that is capable, keeps data on hardware they control, can be reached from anywhere and runs tasks on a schedule.
How it works
The write-up covers one day of setup, so everything here is early and self-reported. On that first day the author says the assistant helped with a security audit and hardening, Tailscale installation and troubleshooting, backups and monitoring, a blog content plan, and a draft of the article itself.
- Access: The OpenClaw dashboard is not on the public internet. It is only reachable from the author's laptop, phone or tablet after each one is authenticated with Tailscale, and every new device has to be approved on the server.
- Hardening: SSH uses keys instead of passwords, and security patches install automatically.
- Monitoring: Every hour the assistant checks for failed SSH logins, unexpected network services, abnormal resource use and OpenClaw audit warnings, and sends an alert if it finds something.
- Backups: At midnight it commits and pushes workspace files (notes, scripts, configs) to GitHub, with commit messages written from that day's activity. The author writes, "I never have to remember to commit again."
- Memory: Conversation logs are saved to dated memory files so they can be referenced later.
Planned uses include drafting blog posts and the Fermenting Solutions newsletter, SEO ranking reports, and work with Threddies. An appendix estimates 2 to 4 hours of setup for someone comfortable with Linux, plus about 30 minutes a month of maintenance. It recommends 8GB of RAM and booting from an SSD.
The result
The author estimates the assistant saved about 3 hours on day one. They say the security audit took about 30 minutes and would have taken more than 2 hours by hand. Running costs are about $30 a month in Venice AI DIEM credits for heavy usage. The author compares that with $500 to $1,000 a month for a part-time assistant, or $200 to $500 a month for a managed AI service.
How to build something similar
- Start with a small always-on machine you control, such as a Raspberry Pi or a spare computer, and install an agent gateway like OpenClaw on it.
- Put the machine behind a private network such as Tailscale before you do anything else, so the dashboard is never exposed to the public internet.
- Give the agent two scheduled jobs first: a nightly backup of its workspace to a private Git repo and a regular security check that alerts you.
- Track what you spend on model credits for the first month so you know your real cost before adding more tasks.
The original post
Want to build something like this?
Join the Shipping Skool community free. Paid members get live bootcamp calls every week.
Join Shipping Skool — FreeMore use cases
- How Vinay Moved an AI Second Brain from Claude Code to Grok BotPersonal Productivity
- How Evan Kirstel Uses Claude Cowork as a Help Desk for Mac and iPhonePersonal Productivity
- How Khe Hy Built an Ambition-Upgrade Skill for Prompts and TasksPersonal Productivity
- How Sherwood Uses Hermes as a Personal Agent for Inbox TriagePersonal Productivity