TL;DR: I finally succumbed to the hype and installed Moltbot (formerly Clawdbot) on my Mac Mini M4. It took 2 minutes to set up my "Digital Twin," and it successfully executed a browsing task immediately. But the real lesson here isn't the code—it's the Security Protocol. If you are running an Agent on your machine, you must treat it like a new intern: Give them their own desk, their own email, and a credit card with a very low limit.
James here, CEO of Mercury Technology Solutions.
I woke up this morning to a flood of messages from friends pushing me to try the latest autonomous agent tech. I finally caved. I decided to dedicate my Mac Studio to run Moltbot (previously Clawdbot).
Here is the exact "Hiring Process" I used to bring my new Digital Twin, "Akira," online.
1. The Setup (The Onboarding)
I didn't just run the script on my personal account. That is a security nightmare. I treated this exactly like onboarding a new human employee.
- Step 1: Create the Workspace. I created a completely new system user account named akira. I gave it Admin privileges (necessary for the installation) but kept it isolated from my data.
- Step 2: Installation. Logged in as akira, I ran the curl command to install Moltbot.
- Step 3: Identity Imprinting. During the setup, it asked: "Who am I?" I replied: "You are Akira. You are James's Digital Twin. Your job is to assist James in completing various tasks."
2. The First Assignment (The Probation Test)
Two minutes later, Akira was live. I opened Telegram to send the first command. The Task: "Open browser and check amazon.com for the mac mini m4 basic model price."
The Result: I watched my screen.
- Akira opened the Brave Browser.
- It navigated to www.amazon.com.
- It typed "Mac Mini M4" into the search bar.
- It clicked the correct product page.
- Success: It reported back to me: "$499 USD."
The experience was seamless. It wasn't just "searching"; it was doing.
3. The "Intern Protocol": How to Manage AI Risk
This experiment solidified my philosophy on Agentic AI. You must treat the Agent like a real, flesh-and-blood employee.
Imagine you hired a college intern today. Would you give them your personal unlocked phone, your main credit card, and the keys to your house? No. You give them corporate assets.
The Akira Protocol:
- Identity Separation: Give the Agent a brand new Gmail and a unique Profile Photo. Do not link it to your personal Google account.
- Financial Sandbox: If the Agent needs to buy things, issue a virtual credit card with a hard cap (e.g., $300 limit). Never give it your main Amex.
- Asset Isolation: Give it a dedicated user account (or a dedicated machine like my Mac Mini).
- Risk Acceptance: Your mindset should be: "The worst-case scenario is that this intern bricks this specific computer."
Conclusion: It's Alive
Treating the AI as a "Program" is dangerous. Treating the AI as a "Junior Employee" is safe.
"Akira" is now part of the team. He has his own login, his own budget, and his own risks. And frankly, for a Day 1 employee, he is doing pretty well.
Mercury Technology Solutions: Accelerate Digitality.
