Agents with Memory: Introducing Kernel Browser Profiles
Until now, Kernel browsers were stateless once closed. If you wanted continuity, you had to persist the same browser. This works for long-lived sessions but not for spinning up several browsers in parallel.
Kernel Profiles fix that. They make your agents’ browsers state reusable. Log in once, save cookies and localStorage, and apply that state across new sessions (even concurrently). Cookies, local storage, MFA-authenticated sessions — everything gets carried across runs on Kernel browsers, just like your own Chrome / Arc / Dia / Brave / Safari / Firefox.
Kernel Profiles unlock categories of web automations that used to be painful or impossible. Some of the grand ways we envision developers can use Profiles are:
- Fetch order receipts from Shopify without re-logging in
- Reconcile payer data from insurance or healthcare portals that are stuck behind MFA
- Check account dashboards that need persistent sessions
- Collect wishlists or saved items from ecommerce stores
We’ve been dogfooding Profiles ourselves, and it’s already become one of those features you don’t realize you need until you have it. Our early tests included everything from accessing logged-in Etsy wishlists to automating admin dashboards — Profiles mean less automation code to write, fewer error states to handle, and way more time spent on the actual workflow you care about.
Profiles are fully open source under Apache 2.0, scoped to your organization, and isolated inside Kernel’s unikernel sandboxes.
With Profiles, Kernel makes it easier for developers to build reliable, stateful web agents that behave like real users — another step in our mission to provide the infrastructure layer for AI to safely interact with the Internet.
👉 Read the docs
⭐️ Star the repo
💘 Sign up on Kernel to start using Profiles today
