B4n1Web is an ultra-lightweight agentic browser engine (6MB) for AI agents. It provides three browsing modes: Light (HTTP fetch + HTML parsing), JS (JavaScript extraction), and Render (full headless browser with Chromium).
Perfect for web scraping, data extraction, E2E testing, and autonomous AI agents that need to navigate and extract content from websites without the overhead of a full browser.
Features:
- Ultra-lightweight binary (6MB)
- Python SDK available on PyPI
- MCP server for integration with Claude, Cursor, OpenCode, Windsurf, Antigravity
- SQLite-based SecurityShield for safe browsing
- Multiple modes: Light, JS, Render
MCP Server:
Command: b4n1web mcp -p 8080
Port is configurable — change 8080 to any available port
Tools: goto (navigate & extract content), get_links (extract all links from current page)
Overview
B4n1Web - Ultra-lightweight Agentic Browser Engine
Installation
Binary (6MB) curl -sL https://github.com/B4N1-com/b4n1-web/releases/latest/download/b4n1web-v0.6.2-flat.tar.gz | tar -xz && ./b4n1web --version Python SDK pip install b4n1-web
Quick Start
from b4n1web import AgentBrowser, BrowserMode browser = AgentBrowser(mode=BrowserMode.LIGHT) page = browser.goto("https://example.com") print(page.markdown) print(page.links) browser.close()
MCP Server
b4n1web mcp -p 8080 Then connect via your MCP client. Port is configurable.
Features
- Ultra-lightweight binary (6MB)
- Python SDK available on PyPI
- MCP server for integration with Claude, Cursor, OpenCode, Windsurf, Antigravity
- SQLite-based SecurityShield for safe browsing
- Multiple modes: Light, JS, Render
Modes
- Light: HTTP fetch + HTML parsing (~6MB)
- JS: Light + JavaScript tag extraction (~6MB)
- Render: Full headless Chromium with screenshots
Tools
goto(url, mode="light")- Navigate and extract contentget_links()- Get all links from current pageget_screenshots()- Get screenshot in render mode
Server Config
{
"mcpServers": {
"b4n1web": {
"command": "b4n1web",
"args": [
"mcp",
"-p",
"8080"
]
}
}
}