What if you could have your own AI assistant running around the clock, no expensive hardware, no upfront investment, for less than the cost of a streaming subscription?
That’s exactly what I built using Amazon EC2 and OpenClaw. Below is the step-by-step process I used to get it running.
Six Steps:
1. Launching an Amazon EC2 Instance
The foundation of this setup is a cloud server from AWS. EC2 lets you spin up a virtual machine in minutes, and if things don’t work out, you’re not stuck with expensive hardware.
Create an AWS Account: Head to the Amazon EC2 console and sign in. If you don’t have an account, it takes about two minutes to create one and there’s a free tier to start.
Launch an Instance: Search for “EC2,” click “Launch Instance,” and name it something like “open-claw-demo.”
Select Your Setup: Choose Ubuntu as your operating system and c7i.large as the instance type. This gives you enough power to run OpenClaw smoothly without breaking the bank.
2. Configuring Storage
The default storage is only 8 GB, which isn’t enough. Bump it to 20 GB as a starting point, 30 GB for regular use, or up to 100 GB if you’re running heavy workflows and storing a lot of files.
3. Security & Access Setup
Create a Key Pair: Click “Create new key pair,” select RSA, and choose .pem for Mac/Linux or .ppk for Windows. This file is the password to your server, so store it somewhere safe.
Network Settings: Allow SSH traffic, then add a Custom TCP rule with the source set to “Anywhere” and the port range set to 8789. This opens the port for the OpenClaw web interface.
Launch & Connect: Hit launch, navigate to your instance, click “Connect,” and a terminal opens right in your browser. No extra software needed.
4. Installing OpenClaw
With the server running, installation is a single command. Go to openclaw.ai, copy the install command, paste it into your EC2 terminal, and hit enter. It takes a minute or two. Once it’s done, OpenClaw is running 24/7 on your cloud server.
5. Onboarding & Model Selection
Quick Start: During onboarding, select “Quick Start” to get up and running fast.
Choose Your Model: I went with Anthropic and selected Claude Opus 4.5, which has been incredibly impressive for complex tasks. You can also use OpenAI, Moonshot AI (Kimi K2.5, which offers free credits), or several other providers.
Choose Your Channel: OpenClaw lets you talk to your AI through Telegram, WhatsApp, Discord, Slack, Signal, iMessage, and more. You can also just use the terminal or the built-in web interface.
6. First Conversation & Memory Setup
The final step is “hatching” your bot. Select “Hatch in the TUI” and your OpenClaw comes online with a clean slate. It will ask you questions: Who am I? What should I call myself? Who are you?
This is the time to brain dump everything: your goals, your workflows, your preferences. The more context you give it in these initial sessions, the more useful it becomes. Everything gets saved to memory, and from here your AI starts working for you.
From here, there’s a ton you can do: adding skills, configuring hooks, organizing memory, and building custom workflows. I’ll be covering all of that in upcoming videos.
A great way to see what it is capable of is just by asking it. Have fun!
