Each customer's logged-in browser is a service your agent calls

You're building an AI assistant that acts on behalf of paying customers: books their flights, files their expenses, replies to their LinkedIn messages. Each customer's logins, cookies, 2FA-signed sessions, and saved cards live somewhere. The wrong place is "all in one browser, partitioned by clever code."

The right place is each customer's container, each with its own Hoody Browser instance, each with its own URL. Customer #847's session state is in container-847's filesystem. The agent dispatching her tasks knows her browser as https://{projectId}-{containerId}-browser-1.{serverName}.containers.hoody.com. That's the address.

A thousand customers, a thousand mostly-idle containers on one box: near-identical browsers whose disk blocks are shared copy-on-write and whose binaries the kernel caches once. When customer #847 asks for something, the agent calls her browser's URL. The other 999 never see her cookies; her cookies are not in their filesystem.