Mmojo.net

Human-first Generative AI.

Mmojo Agent Appliance

AI agents are not quite the panacea everyone hopes, yet. But it is time to get your hands dirty with them — safely and deliberately. I would like to help you start your journey with agents and OpenClaw. I would like to transform your Mac Mini (M1, M2, or M4) into a Mmojo Agent Appliance. Below, I’ll tell you why, then how I can transform your Mac Mini for US$250 (introductory price). You will make that up in saved cloud token costs quickly. My qualification to help is that I am deeply involved in how all the pieces work and deeply concerned about your privacy and security. Questions? Email me.

-Brad

Pain Points with OpenClaw

It’s 2026. Everyone wants to automate their businesses with OpenClaw. When I talk to people who are doing it, I see three pain points while getting started.

1. Security and Privacy

The stories are legion of eager beavers who deleted all their email or overshared their data with OpenClaw. You don’t want to be insecure from the start by flubbing the installation and basic setup. By the way, sending your agent requests to a cloud LLM provider like Anthropic or OpenAI is surrendering your privacy. You’re making a huge step toward your security and privacy just looking at the Mmojo Agent Appliance.

2. Cost of Tokens

OpenClaw requires a connection to a large language model (LLM) to work. Most people use their Anthropic or OpenAI account. Most of these accounts are metered, meaning users effectively pay per token. It is typical for users to spend $50 in tokens just getting basic things set up and trying some experiments, then upwards of $100/day having their agents do routine tasks in “production”. That gets expensive.

3. Wrong Hardware

If you could run the LLM locally on machines you own and control, you won’t incur token costs. Developers with high-end hardware did this almost immediately when OpenClaw burst on the scene. Several YouTubers bragged about assembling $40K Mac Studio clusters to run very large LLMs to support OpenClaw. Meanwhile, people like me submitted bug reports and fixes to help make OpenClaw run using smaller models. This would let the LLMs run privately on more modest hardware.

4. Set Up is Technical

OpenClaw is based on software called the Pi coding agent. Coding agents are for software developers. You need to be comfortable with the command line, networks, ports, services, and the like to get OpenClaw running. Also, you get to set up an LLM server, like Mmojo Server. You’d like everything to start automatically when you power on your appliance.

Considerations

First, some good news! In mid-March, OpenClaw could run using Mmojo Server and the Qwen3.5 9B model without any hiccups. This means your entire system can run on modest consumer hardware you might already have in your home or office.

Let’s list your considerations and solutions:

  1. Keep your personal and business data private and secure.
    • Run OpenClaw on a separate “OpenClaw-only” computer.
    • No data sent to the cloud unless you allow OpenClaw to send it to the cloud.
  2. Zero cloud token spend.
    • Certainly, not $1000+ per month! That’s a car payment.
    • You need a local LLM. You will need to set one up.
  3. Connecting to messaging platforms.
    • You will need a Mac to connect to iMessage.
  4. Expert setup by me.
    • I make a popular private, local LLM server. That requires some expertise.
    • I’ve been working to make Mmojo Server work well with OpenClaw on hardware we can all afford since it was Clawdbot, then Moltbot, then OpenClaw, i.e. from the very beginning of the phenomenon.
    • I do not hand any of the process off to anyone. That’s why it’s $250 and not $49.95. I make sure your appliance is set up securely so you can wade into OpenClaw safely. You will save this money in cloud tokens if you ever do anything serious with OpenClaw.

Mmojo Agent Appliance — Mac Mini

Mac Minis have been a popular way to run OpenClaw, primarily for (2) and (3) to this point. To address (1), you can build a self-contained system running OpenClaw and Mmojo Server.

I would be happy to transform a Mac Mini into a Mmojo Agent Appliance for you. I will install and configure Mmojo Server, install and configure OpenClaw, and have it ready to plug into an Ethernet port on your home router.

  • Price: $250, paid up front.
  • Delivery estimate: 2 weeks from receipt of your Mac Mini.
  • 50 United States only. Other destinations available with prior arrangement to cover tariffs and shipping.

Process

First, we will do this:

  • We will have a quick Zoom call to go over process, contact info, addresses, your Mac Mini, and timelines.

Then, you will do this:

  • You will make sure your Mac Mini has an M1, M2, or M4 processor with at least 16 GB RAM. 24 GB is better. Mmojo Server (for LLMs) does not run fast enough on older Mac Minis with Intel CPUs.
  • You will back up your Mac Mini.
  • You will reinstall macOS, according to Apple’s instructions:
  • You will ship me your Mac Mini and its power cable, preferably in its original box. (I will provide a shipping address when you order.)
  • Alternatively, you can order a new Mac Mini and have it delivered directly to me.

Upon receipt of your Mac Mini, I will do this:

  • I will actually reinstall macOS again to ensure that I start with a clean system.
  • I will install and configure the latest Mmojo Server and the latest OpenClaw for basic operation.
  • I will print a build sheet with important installation instructions and security info.
  • I will ship your Mac Mini back to you. Shipping to the 50 United States with insurance is included in the service price.

When you receive your Mac Mini, you will do this:

  • You will plug your Mac Mini into power and into your home router. You should not need to hook up a keyboard, mouse, or monitor.
  • You will connect to it from another computer.
  • You will configure OpenClaw with service keys, and train agents to do what you want them to do.

Ready to Order?

I accept payment on PayPal.

cards
Powered by paypal

There is an option at checkout for a 1 hour OpenClaw tutorial session with me. It’s an additional US$100. We’ll create two periodic information queries that are reported each time you chat with your OpenClaw agent.

Want a Free Consultation Before Ordering?

Email me with three good 20-minute time slots in the next 3 days, 8am to 8pm US Pacific Time. I will schedule a Zoom call.

Frequently Asked Questions (FAQ)

Nobody actually asked these, but these answers need to go somewhere.

  1. Will a Mac Mini with an Intel CPU work?
    No. The on-board GPU in Apple Silicon (M1, M2, M4) is needed for Mmojo Server to run fast enough to keep up with OpenClaw.
  2. How about MacBook Air, Pro, Neo?
    If you promise me it will be a dedicated machine, just portable, sure. It’s not a good idea to mix your daily driver laptop with an OpenClaw installation. Security and privacy of your data, you inadvertently breaking OpenClaw or Mmojo Server, etc. MacBook Neo doesn’t have enough RAM.
  3. How about AMD Ryzen™ AI Max processors like in a Framework Desktop?
    Of course! Most OpenClaw enthusiasts don’t know about these. Email me if you’re interested.
  4. How about an NVIDIA Orin NX or a cluster of them?
    Of course! Most OpenClaw enthusiasts don’t know about these. Email me if you’re interested.
  5. Can OpenClaw really work with a small model, i.e. 9B parameters?
    Yes, the OpenClaw crew fixed the connection and timeout problems in mid-March. They are making the system prompt smaller. Smaller, more terse prompts with fewer contradictions work better for agents with LLMs of any size, so that trend will continue. That said, Qwen3.5 4B doesn’t work well, and I don’t expect they’ll reach models that small.
  6. Brad, how did you write this offer and not mention drift?
    Well, it’s something the cloud LLM users will surely experience soon enough. Their agents work, then something in the cloud LLM changes, and… Sad trombone sound. Their agents break, or act differently. When you use a private LLM and you don’t change the LLM, there is no LLM drift.
  7. Tell us more about security.
    Security is NEVER bolted on. It is built up layer by layer by not introducing security holes. You will be using a system where an LLM can decide to publicize any data on the system. So do not ever put data that you don’t want accidentally shared on the system. Start thinking about security now.