When I first launched Jarvis, my plan was simple: someone signs up, I SSH into a server, set everything up manually, and send them the login details. Easy enough for the first few customers. But then I asked myself — what if 10 people sign up in one day? Am I really going to sit there provisioning servers one by one?
The answer was obvious. I needed to automate the entire process. So I told my AI agent Tim to build a fully automated provisioning pipeline — from payment to a ready-to-use AI server, without me touching anything.
The Problem With Manual Setup
Setting up a single customer's server manually takes about 30-45 minutes. Here's everything that needs to happen:
- Create a new VPS in the cloud
- Wait for the server to boot
- SSH in and update the system
- Install the AI agent and all its dependencies
- Configure domain, SSL, and firewall
- Create the customer's account
- Send a welcome email with login details
For 1-2 customers a day, maybe that's manageable. For 10? 20? I'd spend my entire day doing nothing but server setup. It doesn't scale.
And there's a bigger problem: timing. If someone signs up at 3 AM, they have to wait until I wake up before they can use anything. That's a terrible customer experience.
The Auto-Provision Pipeline
I told Tim: "Make everything automatic. From the moment a customer pays to the moment their server is ready. I don't want to touch anything."
Here's the 6-step pipeline Tim built:
Step 1: Customer pays through Stripe. The Jarvis website has a pricing page where customers choose their plan and pay. Stripe handles the payment — it's the same payment infrastructure used by companies like Shopify and Amazon.
Step 2: Stripe sends a webhook. The moment payment completes, Stripe fires a webhook event to our system: "This customer just paid for this plan." The provisioning process starts immediately.
Step 3: VPS created via Hetzner API. The system calls the Hetzner Cloud API to spin up a new VPS. It selects the right specs based on the customer's plan and deploys from a pre-built snapshot. The server boots in seconds.
Step 4: SSH runs the provision script. Once the server is up, the system automatically SSHes in and runs a provision script. This installs the latest AI agent, all tools, the memory system, and every configuration file. No manual setup required.
Step 5: Domain + SSL configured. The system creates a subdomain through the Cloudflare API and sets up an SSL certificate automatically. The customer gets a clean URL with HTTPS ready to go.
Step 6: Welcome email sent. Finally, the system sends a welcome email with all the login details. The customer clicks the link, logs in, and starts using their AI agent immediately.
All 6 steps happen in about 2 minutes. I don't touch a single thing.
Not Just an Empty Server
This is the part I'm most proud of. The auto-provision system doesn't just create an empty box. Every server comes fully loaded and ready to use:
- AI agent with persistent memory — it remembers conversations across sessions, like a brain that never forgets
- 18+ pre-built skills — from accounting to content creation to data analysis
- Full tool suite — internet browsing, image generation, code execution, automation
- Chat UI — a clean web interface for talking to the AI through any browser
Customers log in and start talking to their AI immediately. No technical knowledge required. No setup on their end.
Why This Was Built in a Single Session
If I had to write all of this code myself — learning each API, handling error cases, testing the entire flow — it would have taken days. Maybe a week.
But because I have Tim (my AI agent) with real access to the server, he did everything: wrote the Stripe webhook handler, integrated the Hetzner API, wrote the provision script, tested the full flow, and deployed it. All in one session.
This is the difference between an AI agent and a regular chatbot. ChatGPT can tell you how to do this. Tim actually does it — writes the code, runs it, tests it, deploys it.
The Results
Since building the auto-provision system:
- Customers get their server instantly — no waiting for business hours. Sign up at 3 AM, get your server at 3 AM.
- I do nothing — the system handles everything from payment to welcome email.
- It scales infinitely — 100 customers signing up at once? No problem.
- Zero human error — every server is configured identically. No forgotten steps.
It's like having a DevOps team that never sleeps, never takes breaks, and never makes mistakes. The system runs 24/7 without supervision.
The Lesson
The biggest takeaway from this project: if you're doing something manually more than twice, automate it.
I used to think "it's just a few customers, I can handle it manually." But the moment I automated provisioning, I stopped thinking about it entirely. That mental space — not having to worry about server setup ever again — is worth more than the time saved.
And this is what AI agents actually do. Not just answer questions. Not just generate text. They build real systems that run your business while you focus on what matters.
If you want your own private AI server — fully provisioned, ready to use, no technical setup required — that's exactly what Jarvis is. The auto-provision system I described in this article will set everything up for you in minutes. You just log in and start working with your AI. See how it works →
— Pond