Want to try out Amazon Lightsail, but nervous about the technical aspects? I’m going to teach you how to create an OpenClaw instance even if you have zero AWS experience! All you need is an AWS account!
In this tutorial, we’ll learn how to:
- Set up an OpenClaw instance using a “click-to-launch” pre-packaged application
- Pair with your browser and configure OpenClaw using SSH
- Open your OpenClaw dashboard and start chatting with your agent/assistant
- How to take your OpenClaw instance down
You do not need to have any prior knowledge with Amazon Web Services to follow this tutorial!
Looking for more beginner-friendly tutorials on Amazon Lightsail on this blog? Hop over to /lightsail!
Table of Contents
What is OpenClaw?
OpenClaw is an AI-powered chat agent that provides a private, self-hosted (this means that it can technically also run on your computer) AI assistant to perform many tasks autonomously on your behalf.
You can “chat” with your assistant in many ways, including your browser, Telegram, and WhatsApp (you can check out all of the applications it can integrate with here).
Most “chatbots” we are familiar with can “talk to you.” But this chat agent has “eyes and hands,” which means it can read and write files, browse the Internet, run commands, fill out forms, and execute scripts. In a nutshell, it can actually do things.
Be careful, as it is powerful, but may not have the “human logic” of what’s right and wrong that you and I may have as humans (Twitter was full of funny- and sometimes cringe-worthy- ways OpenClaw agents wrecked havoc on people’s systems/life while executing tasks in very “technical” ways).
Used correctly, OpenClaw could be that personal assistant we all dreamed of!
Today, we will be setting it up and hosting it on Amazon Lightsail using a click-to-launch application. Their official website is at: openclaw.ai.
Setting up an OpenClaw instance on Lightsail
Create an OpenClaw instance on Amazon Lightsail
- Go to Amazon Lightsail console, and choose “Create instance”
- Choose your Instance location and instance image:
- Instance location: default (or whatever you want it to be)
- Platform: Linux/Unix
- Blueprint: OpenClaw (Lightsail)
- Optional settings: leave them alone (unless you want to change them)
- Instance plan: General purpose
- Network type: Dual-stack
- Size: for testing purposes, $12 USD/month (first 90 days free- be careful, your use of AI is not free!)
- Choose a name for your OpenClaw instance, and choose a tag (or 50) if you are inclined, and click “Create instance“
- You will be sent directly to your brand new OpenClaw instance dashboard! When the page loads, your instance will probably show as “starting…,” but if you refresh once, you should be ready to get started! (It’s ready when it says “Instance status: Running”.)
Congratulations! You now have OpenClaw running on Amazon Lightsail!
Pair your browser with OpenClaw using browser-based SSH
Now, we need to create secure connection between your brand new OpenClaw instance and your browser by “pairing” your browser with OpenClaw.
No worries though. This is all copy and paste and a few yes/no prompts!
- Go back to your OpenClaw dashboard, and click on “Connect using SSH” under “Access your OpenClaw dashboard“
- A screen will pop up with in-browser SSH terminal; look for information under “OpenClaw Dashboard Access“:
- Dashboard URL (copy it)
- Access Token (take note of it)
- Don’t close this SSH terminal window!
- Take the dashboard URL you just copied, and paste it into a new browser window, and enter the “Access Token” from your SSH terminal into the “Gateway Token” field, and click “Connect“
- You will get a red message that says “pairing required“
- Go back to your SSH terminal, and find the message:
Continue with browser device pairing? (y = pair now, n = skip):- Type in “y” to pair and let the script run
- You will next find the message:
Pending device request: [numbers and letters] Action? (a=approve, r=reject, s=skip all):- Type in “a” to approve the pairing request
- Go back to your OpenClaw gateway dashboard, and click “Connect” again; this time, your OpenClaw dashboard should load!
Huzzah! Your OpenClaw agent is alive! (Alive?) If you need to pair additional browsers, you can do these steps again!
Enable Bedrock API access with AWS CloudShell terminal
If you go to “Chat” on your new OpenClaw gateway dashboard and type “Hello,” you won’t get a response. That’s because the AI capabilities aren’t available to you yet. To actually begin using your chat agent, you need to enable API access for Amazon Bedrock.
We will do this through the AWS CloudShell terminal by running a script. Don’t worry! This is all built-in to the Lightsail dashboard as well!
- Go back to your OpenClaw instance dashboard on Amazon Lightsail, and scroll down to “Enable Amazon Bedrock as your model provider“
- “Copy the script” and click on “Launch CloudShell“
- Once you see
~$, paste in the script, and click enter, and confirm that in one of the last lines, it saysDone
The script:
- Created an IAM role for your OpenClaw instance
- Attached a policy to grant access to Amazon Bedrock APIs
- Attached a policy to grant access to AWS Marketplace permissions
- Configured the instance profile to use this role
If this is your first time using Anthropic models with Amazon Bedrock, you will need to complete “First Time Use (FTU) form” to gain access. Please refer to this documentation to learn more: request access to models.
To access the form, go to Amazon Bedrock console, navigate to “Model catalog,” and select Claude Sonnet 4.6 and fill out the form.
Start a conversation with your new AI assistant!
Go to the “Chat” window in your OpenClaw dashboard, and start a conversation with your OpenClaw assistant!
Take down your OpenClaw Lightsail instance
With the many ways running OpenClaw can charge you on AWS, it’s very important that you are mindful of your usage and stop or delete the instance when you are done with it.
Thankfully, it’s very quick an easy: go to the OpenClaw dashboard, and click Delete.
If you want to keep your data, create a snapshot of the instance before deletion.
Other things to consider with OpenClaw on Lightsail
Looking for an in-depth introduction and tutorial on setting up OpenClaw on Lightsail (especially the technical specifications)?
Check out the official documentation and Frequently Asked Questions on AWS’s KB.
Costs associated with running OpenClaw on Lightsail
AWS’s official documentation provides information on costs associated with running OpenClaw on Lightsail. There are a few ways you will be charged. (Information pulled from: here.)
- Lightsail instance — You pay for the instance plan you selected (e.g. the 4 GB plan). Lightsail plans are billed on an on-demand hourly rate, so you pay only for what you use. For every Lightsail plan you use, we charge you the fixed hourly price, up to the maximum monthly plan cost.
- AI model usage (tokens) — Every message sent to and received from the OpenClaw assistant is processed through Amazon Bedrock using a token-based pricing model. Costs vary by model — some models are more expensive per token than others.
- Third-party model subscriptions — If you select a third-party model distributed through AWS Marketplace (such as Anthropic Claude or Cohere), there may be additional software fees on top of the per-token cost. These appear as separate line items under AWS Marketplace in your bill.
- Data transfer overages — Each Lightsail plan includes a monthly data transfer allowance. If your OpenClaw instance sends or receives more data than your plan includes, overage charges apply for data transfer out.
- Snapshots — Manual and automatic snapshots of your Lightsail instance are billed based on the amount of storage used.
One thing you need to be cognizant of is that your charge to use Amazon Lightsail to run OpenClaw is separate from your usage charge to use the AI assistant capabilities through Amazon Bedrock. So each interaction you have with your assistant will be costing you tokens (which, in turn, is money).
Connect to a messaging channel
You can also connect your OpenClaw to “talk” to you via messaging applications like Telegram and WhatsApp. You can follow the instructions on AWS’s official documentation to set this up: Connect a messaging channel.