Scalebrowser

Scalebrowser is browser infrastructure for AI agents that need to operate real websites while staying signed in between runs. Instead of renting a remote browser from a cloud service, a team installs Scalebrowser on its own Windows machine and lets an agent drive persistent profiles there. Each profile keeps its own cookies, storage, logins, device identity, and proxy settings, so recurring work can resume as the same browser identity rather than starting from a blank session every time.

The product has two main pieces. A local daemon creates profiles, starts and stops browsers, enforces launch checks, and exposes the interfaces that agents and automation code use. A separately delivered Chromium fork is the browser engine. Scalebrowser builds and signs that engine, and the daemon verifies its signature before launch. Profiles are headless by default but can open visibly when an operator needs to watch or intervene. All profile data, installed engines, and the local database stay on the machine running the daemon, and there is no hosted mode. A bundled management app gives the operator a visual place to create profiles and inspect one before handing it to an agent. The whole data directory can be backed up to preserve profiles, while uninstalling the desktop app leaves that data in place for a later reinstall.

Persistence goes further than keeping a cookie jar. A deterministic seed gives each profile a consistent set of browser and device values across restarts, while isolated storage keeps one profile's sessions away from another. Stored passwords are encrypted on disk and entered into the relevant site without exposing the value to the agent. Profiles can also retain passkeys and receive email confirmation codes through an inbox assigned to that profile. Sensitive values discovered on a page can reach the agent as placeholders instead of raw secrets. A human input layer produces clicks and keystrokes, and the product includes handling for several kinds of verification challenge. Detailed run trails record tool calls, results, and screenshots so an operator can review what happened.

Teams can connect an agent through MCP, use the Python or Node SDK, manage profiles through a REST API, or attach existing automation over the Chrome DevTools Protocol. Those paths all talk to the same local daemon and the same profiles. Direct CDP is useful when a team already owns the browser logic, while MCP gives a coding or chat agent a smaller browser tool set. In the documented MCP flow, the agent takes a lease on a profile, opens a page, receives a structured text map, and releases the profile when the work is done. The REST API is aimed at profile and fleet management rather than page interaction. Documented client setups include Claude, Cursor, Codex, ChatGPT, GitHub Copilot, and other MCP capable tools.

Scalebrowser fits developers and automation teams whose agents must work inside authenticated web products, tools without APIs, or many separate account identities. It can support repeated research, marketplace operations, account administration, and other browser work where losing a session would break the workflow. Groups, presets, bulk actions, profile synchronization, extension support, and proxy controls cover the move from one browser to a larger fleet. The local design is especially relevant when credentials and profile state shouldn't live in a browser vendor's cloud. Operators still need to decide which sites and accounts an agent is permitted to use, and the product doesn't remove the need to follow each site's policies.

Its most distinctive choice is to treat the browser identity as durable infrastructure on hardware the operator controls. That offers more continuity than launching stock Chromium with a fresh profile, but it also creates setup responsibilities. Scalebrowser currently ships for 64-bit Windows 10 and 11, requires a real GPU to launch profiles, and recommends a suitable residential or mobile proxy for identities that shouldn't leave through the machine's normal address. A profile may run on only one machine at a time, which protects state but matters when scheduling work across a fleet. There is no managed cloud fallback, and the installer currently triggers a Windows SmartScreen warning because it doesn't yet have a Windows publisher certificate.

Access is paid, with a seven-day trial that requires a card. The Start plan is listed at €19 per month for one concurrent browser on one machine, while higher plans raise browser concurrency and machine limits. Every plan includes unlimited profiles and the same core product, so pricing is based mainly on how much can run at once. Solo is listed at €39 for five browsers, Studio at €99 for twenty, and Fleet at €249 for fifty across ten machines. Published add-ons can extend Fleet without a sales call. That makes the entry plan practical for testing a single persistent workflow, while larger fleets become a meaningful monthly infrastructure cost.