Amazon Web Services (AWS) said on March 4 (local time) it will officially support OpenClaw, an open-source AI agent tool, on Lightsail, its virtual private server (VPS) service.

OpenClaw is an open-source platform that lets users run a personal AI agent directly on their own device.

It runs AI agents through a browser and connects to messaging apps such as WhatsApp, Discord and Telegram to perform tasks including managing email, web searches and organising files. It uses an agent approach that autonomously handles tasks beyond simply answering questions.

There have previously been cases of running OpenClaw on Amazon EC2 instances. However, installation was complex and security settings were burdensome, making it hard for general users to access. AWS said the Lightsail integration is intended to address these issues.

The company said users can select an OpenClaw blueprint when creating an instance in the Lightsail console. A 4GB memory plan is the recommended specification. After creating the instance and completing browser pairing in an SSH terminal, users can use the AI assistant immediately without additional settings.

The default AI model provider is set to Amazon Bedrock. A single script run can enable Bedrock API access permissions. Third-party models such as Anthropic Claude and Cohere can be used through AWS Marketplace for an additional fee.

Costs are charged at the hourly on-demand rate based on the selected instance plan. A separate Bedrock token-based fee applies for AI response processing.

On security, AWS recommended not exposing the OpenClaw gateway to the public internet. It instructed users to regularly replace the gateway authentication token and not to write it directly in the configuration file.