Scalebrowser

Scalebrowser provides self-hosted browser infrastructure so AI agents can run browser tasks reliably over time. Instead of starting from a fresh session each run, you assign each agent its own isolated profile with a persistent identity derived from a profile seed—so sites see the same visitor across restarts.

Each profile has its own logins, cookies, and device-related state, with encrypted stored credentials and per-profile inbox delivery for verification codes. Two-factor codes are computed inside the profile, and passkeys persist through crashes and window closes. Scalebrowser is built to handle the real friction of sign-in flows, including permission prompts and device pickers, so automation doesn’t stall waiting for manual decisions.

Under the hood, Scalebrowser uses a patched Chromium fork designed for agent-friendly operation. It runs with your machine’s real GPU and screen values, and it includes mechanisms to avoid contradictory profile states starting. For agent execution, it supports multiple connection modes—MCP, a REST API, SDKs for Python and Node on direct CDP, and direct CDP for existing browser-use setups.

When agents interact with pages, Scalebrowser turns each tool call into observable run steps with results and screenshots. It also includes captcha solving for live captcha widgets and offers features like limits and secret filtering, plus compact page mapping/deltas, to keep runs controlled and easier to understand.

Pricing and trial availability: Scalebrowser offers a 7-day free trial.

What problem does Scalebrowser solve for AI agent login sessions?

It keeps agent browser profiles persistent so logins, cookies, and passkeys don’t reset on every run. Each agent can stay signed in with its own isolated identity.

How does Scalebrowser keep an agent’s identity consistent across restarts?

Each agent uses an isolated profile whose identity is derived from a profile seed, and that identity survives restarts. Sites should see the same visitor rather than a new device.

What authentication steps can Scalebrowser handle inside a profile?

It supports full sign-in flows, including permission prompts and device pickers. It also handles real verification steps like email/SMS codes and 2FA, with two-factor codes computed inside the profile.

How do stored credentials and 2FA secrets work?

Passwords are stored encrypted on disk and typed into forms by the profile, so the agent doesn’t directly see the value. Two-factor codes are generated inside the profile from a secret that doesn’t leave it.

How can my agent connect to Scalebrowser?

You can connect over MCP (HTTP or stdio), via a REST API, or using Python/Node SDKs on direct CDP. If you already run browser-use, you can connect directly via CDP.

Does Scalebrowser provide visibility into what happened during a run?

Yes. Each tool call becomes a run step with results and a screenshot, and you can use page mapping/deltas plus run trails. It also includes limits and secret filtering to help keep runs controlled.