Project MonetRequest demo
Home/Blog/BrowserSkill: Tencent's Logged-In Browser Tool for AI Agents

Automation · Project Monet Briefing

BrowserSkill: Let AI Agents Use Your Logged-In Browser

Tencent's BrowserSkill bridges shell-capable AI agents into a real logged-in Chromium profile through a local CLI, daemon and extension.

Published 2026-09-01 · Updated 2026-09-01 · By Project Monet Editorial Team

AI agent connected through the BrowserSkill CLI and local daemon to a logged-in Chromium Agent Window.

01

What BrowserSkill is

BrowserSkill is an MIT-licensed open-source bridge from Tencent that lets shell-capable AI agents operate a real Chromium browser using existing login state. Its repository explicitly names Cursor, Claude Code, OpenAI Codex and several other agent harnesses.

  • The agent calls the bsk CLI.
  • A local daemon relays commands over localhost.
  • A Chromium extension performs actions in a separate Agent Window.

02

Why logged-in browser access matters

Many useful browser workflows depend on cookies, account permissions, MFA state or a real application profile. BrowserSkill targets that gap instead of always starting a fresh automation browser.

  • Existing user tabs must be explicitly borrowed when needed.
  • Human takeover is supported for captchas, logins and confirmation steps.
  • Chrome and Edge are the documented browser targets; Firefox is not currently presented as supported.

03

Useful and unsafe use cases

The documented skill covers browsing, page reading, form interaction, screenshots and web-app validation. Logged-in access can also make internal dashboards and authenticated admin tools reachable to an agent.

  • Start with read-only tasks.
  • Keep financial, publishing, access-control and destructive actions behind human confirmation.
  • Do not treat logged-in access as permission to bypass site rules or human-verification mechanisms.

04

Security and privacy boundaries

The feature that makes BrowserSkill useful also increases the consequences of a bad instruction. Review extension permissions and source code, use least-privilege accounts where possible, and keep the agent's scope narrow.

  • The bridge is designed to run locally.
  • Developer privacy disclosures are not an independent security audit.
  • An isolated automation browser remains preferable when deterministic testing or strong separation from production sessions matters most.

Sources

Primary and supporting sources

Facts were rechecked against the linked sources immediately before publication. Pricing, product availability and rollout status can change.

Project Monet

Useful signals. Clear decisions. Better digital work.

Project Monet turns relevant shifts in AI, creator tools and the web into practical context—and builds focused websites for businesses ready to grow.

Request a free homepage concept