If you’ve got a Mac Mini, old MacBook, or really any machine collecting dust, there’s a project worth considering. OpenClaw is a free, open-source AI agent platform that runs locally on your own hardware. It connects to chat apps like Telegram, Slack, or Discord and can handle scheduled tasks, automations, and monitoring in the background while you go about your day.
If you spend any amount of time online, be it YouTube or social media, you’ve no doubt run across ClawdBot as it was first named, then MoltBot before finally settling on OpenClaw. Here’s some searches to get you started if you’ve missed it:
I’ve been running it for about three weeks now on a dedicated Mac Mini and it’s become one of the more useful things in my setup. Here’s what it actually does and why an old Mac is a solid fit for it.
What OpenClaw Does
Think of it less like a chatbot and more like a background operator. You give it scheduled jobs and it runs them. You message it through Telegram or Slack and it responds with access to your files, email, calendar, web search, and whatever else you connect.
Some examples from my setup:
Every morning it sends me a daily briefing with yesterday’s website traffic, local weather, and relevant news from RSS feeds I care about. It ends with a prompt: “What’s the single highest-leverage thing I can ship before lunch?” Simple, but it reframes the morning.
It runs automated prospecting for my business every weekday. Finds one local prospect every morning before I wake up and another every afternoon, verifies the evidence, and appends it to a Google Sheet. If it can’t verify with a real quote and URL, it skips it. No guessing.
It posts my outstanding to-do items to Telegram four times a day so nothing falls through the cracks.
I also set up a second agent specifically for marketing strategy that delivers a weekly memo reviewing my web properties and suggesting priorities.
None of this is magic. It’s just scheduled automations with an AI brain behind them instead of a dumb script.
Why an Old Mac Works
OpenClaw needs a machine that can stay on 24/7 with a decent internet connection. That’s it. The minimum requirements are modest: Node.js 22+, 2GB RAM minimum (4-8GB recommended), and a stable network connection.
That means your old Mac Mini, a retired MacBook running with the lid closed, or even an older iMac you were about to sell for $200 on Facebook Marketplace could be doing useful work instead of sitting in a drawer.
I’m running mine on an entry level M4 Mac Mini, but the software would run fine on an M1 or even an older Intel Mac. The AI processing happens in the cloud via API calls. Your local machine is just the orchestrator, not the engine.
If you’ve already read 10 Creative Ways to Repurpose Your Old Mac Mini, this is another solid option for that list. Dedicated always-on hardware for a dedicated purpose.
What You Need to Know Before Starting
It’s not plug and play. You need to be comfortable with the command line. Installation is straightforward:
curl -fsSL https://openclaw.ai/install.sh | bashbut configuration takes time. Plan on spending a weekend getting it dialed in.
API costs are real but manageable. The software is free. The AI models it talks to are not. You’ll connect it to OpenAI, Anthropic, or both, and pay per token. My costs settled to a reasonable level after some early mistakes. More on that in a second.
Start with one automation. Don’t try to build a whole system on day one. Set up a daily briefing or a simple reminder loop. Get that working reliably. Then add the next thing.
Sandbox it. I gave my OpenClaw instance its own email account and its own Apple ID. It has zero access to my real accounts. This is especially important if you manage client work or sensitive data. Start with read-only access to things and build trust over time.
Lessons From My First Three Weeks
Watch your token usage. I set up a Gmail polling job that ran every 15 minutes during waking hours. That burned over half a million tokens per day before I caught it. Every scheduled task needs a quick sanity check on how often it runs and how many tokens each run costs.
Memory drift is a thing. AI agents forget decisions over long conversations. I solved this by writing everything down in files the agent reads on startup: permanent rules in one file, daily logs in another, current status in a third. If it’s not written down, it didn’t happen.
Let automations be visible. I tried suppressing routine notifications to keep things quiet. Bad idea. I’d rather see my scheduled tasks fire than wonder if they ran. Proof-of-life beats inbox silence.
Config mistakes can crash everything. One wrong key in a config file took my whole system offline. Validate changes before restarting.
Is It Worth the Effort?
If you like building systems and you’ve got hardware sitting around, yes. OpenClaw turns a dusty Mac into something that actively helps you every day. The upfront investment is time, not money.
If you’re expecting Siri but better, this isn’t that. It’s a tool for people who are comfortable getting their hands dirty and want a personal automation layer they fully control on their own hardware.
The M4 Mac Mini starts at $499 new if you want dedicated hardware. But the whole point is that you probably already have something that’ll work.