Tencent /BrowserSkill

Let AI agents use your real, logged-in browser without interrupting your work. CLI + extension for browser automation across any shell-capable AI agent.

★ 9,967⑂ 714👁 25▤ 29
TypeScriptCreated 2026-06-22
+0
Daily
+0
Weekly
+1,246
Monthly
9,967
stars
714
forks
25
watchers
–
prs
–
issues
29
releases

README

BrowserSkill


  



  Your agent. Your browser. One connection.
  Turn the browser you already use into a workspace for your AI agent.



  
  



  



  English · 中文
  Website ·
  Demo ·
  Features ·
  Website debugging ·
  Documentation ·
  Changelog


BrowserSkill connects your AI agent to Chrome or Microsoft Edge, using your existing logins. Give it a task to read, click, fill, or investigate. Work stays visible in a separate Agent Window, and you can take over when needed. Existing tabs can be explicitly borrowed and returned when the task ends.

 See it in action


▶ Watch the demo · 81 seconds · English

https://github.com/user-attachments/assets/6111d978-baa6-4a76-92c4-4f51d47ae617




  Click the cover to expand the video · Download MP4
  Real browser actions on a local demo workspace, recorded from main.


 Bring the agent you already use


  
    Codex
    Claude Code
    Cursor
    OpenClaw
  
  
    CodeBuddy
    WorkBuddy
    Pi
    Hermes Agent
  


These and other agents that can run shell commands use the same bsk CLI. DeepSeek Harness has a dedicated plugin with native browser tools and task previews. You choose the agent and model; BrowserSkill connects it to your browser.

 What you can do

| Your goal | What BrowserSkill provides |
| --- | --- |
| Work where you are already signed in | Read documents, search internal sites, and complete web workflows with the browser's existing login state. |
| Turn instructions into browser actions | Read page text and controls, click, type, fill forms, drag items, select ranges, and manage tabs. Capture viewport or full-page screenshots; upload and download files in local mode. |
| Stay in control of the task | Work alongside the agent with shared control, pause it while you edit, and let it reconcile your changes before continuing. Tasks remain visible in their own window; existing tabs can be explicitly borrowed. |
| Take web data with you | Export loaded tables and lists as JSON/CSV, with column names, page/frame sources, merged-cell metadata, and explicit coverage. Structured extraction |
| Find the evidence behind a failure | Connect actions to requests, response bodies, console messages, and page changes. Inspect performance, slow APIs, and suspected duplicate requests; test a hypothesis with explicit HTTP rules or request replay. Website debugging |
| Keep a replay of the work | Record one authorized task tab as a silent MP4, then preview, save, or delete it from the extension. Task video recording |
| Continue when the page is ready | Wait for a requested element condition with an explicit timeout before the next step. Element waits |
| Reuse what you already observed | Execute one chosen action and synchronize with step; explicitly save the full observation, then search or read it with view without another browser capture. Steps and saved observations |
| Use the right browser for the job | Name browser instances, bind a task to a specific profile, or pair a server-side agent wit

Snapshot: 2026-11-04T14:00:00Z · Data license Apache-2.0 · JSON