Delegate website tasks with local memory and controlled account access
Aside is an AI browser built to carry out tasks across websites where a user already works. It combines normal browsing with an agent that can use page context, account sessions and selected files to complete a requested outcome. The official setup guide currently specifies macOS 15.0 or later and requires an Aside account, making it a desktop browser rather than an extension added to another browser.
A task can begin from the main task page or from a side panel associated with the current page. Users describe the result they want and select permissions or a working folder when needed. The product is designed to interact with websites directly instead of requiring a separate connector for every service. That can support research, administrative work and communication workflows, but the user should still check the affected account and requested action before approving a consequential step.
Aside stores task-related memory locally and makes that memory inspectable and editable. Its password-manager workflow is intended to let an agent sign in through controlled autofill without receiving raw credentials as ordinary model text. The website describes approval gates for sensitive actions, scoped access and logs of credential use. These controls are useful when the same browser contains personal accounts and work systems, because the task needs a narrower boundary than simply having a logged-in session available.
The free plan includes monthly credits and a limited number of routines, with larger plans adding capacity and automation features. Aside also supports connected model subscriptions or user-provided API keys. Local-first storage must not be confused with entirely offline processing: the privacy policy says hosted model providers receive the task context needed to run a request, while selected sync and account features use server-side systems. Users should review those settings before working with confidential pages or files. The practical starting point is a reversible task with a clear result, followed by inspection of the output and permissions before enabling a recurring workflow.
