BB-Browser

BB-Browser turns your web browser into a structured API. Instead of scraping HTML or reverse-engineering APIs, BB-Browser uses site-specific adapters that interact with websites like a human user — and return clean, structured data.

How it works

BB-Browser controls a headed Chrome instance and exposes platform-specific commands through the Pinix Hub:

Agent / CLI
    │
    ▼
Hub ──► bb-browser-daemon ──► Chrome
            │                    │
            │  site adapters     │  real browser
            │  ├─ twitter        │  ├─ cookies
            │  ├─ google         │  ├─ JS execution
            │  ├─ hackernews     │  ├─ DOM interaction
            │  └─ ...            │  └─ anti-detection
            │                    │
            └── structured data ◄┘

Platform coverage

BB-Browser ships with 36 platform adapters and 103 commands:

Category	Platforms
Search	Google, Bing, DuckDuckGo
Social	Twitter/X, Reddit, Xiaohongshu, Weibo
Tech	HackerNews, GitHub, Stack Overflow, arXiv
Knowledge	Wikipedia, Zhihu, Bilibili
Business	Xueqiu (stocks), 12306 (trains), Amap (maps)
Media	YouTube, Bilibili, Douyin
And more	20+ additional platforms

Usage

BB-Browser is automatically started by pinix start. Its commands appear in pinix hub list:

# Search Google
pinix invoke google search "pinix ai agent"

# Get HackerNews top stories
pinix invoke hackernews top

# Search Twitter
pinix invoke twitter search "AI agents" --sort recent

# Get stock quotes
pinix invoke xueqiu quote "AAPL"

Structured output

All commands return structured JSON:

pinix invoke hackernews top --json

{
  "items": [
    {
      "id": "12345",
      "title": "Show HN: Pinix — Capability platform for AI Agents",
      "url": "https://github.com/epiral/pinix",
      "score": 342,
      "author": "builder",
      "comments": 128,
      "time": "2h ago"
    }
  ]
}

JQ filtering

Filter output on the command line:

pinix invoke hackernews top --jq '.items[:3].title'

Authentication

Some platforms require authentication (Twitter, Xiaohongshu, etc.). BB-Browser uses your existing browser cookies:

Open Chrome (the one managed by bb-browser)
Log in to the platform manually
BB-Browser uses those cookies for subsequent commands

Platforms that don’t require login (HackerNews, Wikipedia, Google) work immediately.

Architecture

BB-Browser has three components:

Component	Purpose
`bb-browser`	CLI — for direct terminal use
`bb-browser-daemon`	Edge Clip Provider — connects to Hub, manages Chrome
`bb-viewer`	Stream viewer — for remote browser viewing (auto-downloaded)

The daemon manages the Chrome lifecycle and routes invocations to the appropriate site adapter.