Aloria

Screenshots

Problem Statement

This project uses AI agents to plan tasks, parse images and execute steps of all kinds of tasks by operating on a web browser. Users need to login with their wallet, pay the node that host the service to ask AI agents to complete tasks like booking tickets, monitoring news or price changes, gathering information from reddit, etc. Users can also publish their task prompts to claim ownership. XMTP sends users asynchronous updates about their tasks.

Solution

The app is built with next.js and typescript and we use:React for frontendClaude agents to plan for tasksGemini to parse screenshots of web browser content to decide the most appropriate next interactionPlaywright to operate on web browsersSharp to process screenshots and imagesStory to publish user promptsXMTP to send out agent updates to user walletsDynamics to allow user login with different walletsWalrus to store/retrieve the screenshots of task execution