About B4n1web
B4n1web is a lightweight headless browser engine for AI agents, distributed as a Rust binary with language SDKs. It supports lightweight fetching, JavaScript extraction, and Chromium rendering, along with network interception and safe-browsing controls.
Key features
- 01Light fetch, JavaScript extraction, and Chromium rendering
- 02Extract Markdown, links, and screenshots
- 03Click, type, and wait for page elements
- 04Intercept network traffic and mock responses
- 05Filter domains and enable safe browsing
Use cases
- 01Retrieve page text and links for an agent
- 02Render pages and capture screenshots
- 03Automate page clicks and text entry
- 04Block page resources or mock network responses
MCP support
Its stdio MCP server exposes browser tools for AI agents to navigate pages, extract content, and perform browser actions. The README lists either 33 or 35 MCP tools.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
We haven’t collected a package or remote endpoint for this server. Follow the instructions in the project README.
Tools
Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.
Details
From the project README
🌐 B4n1Web — Agentic Browser Engine
Ultra-lightweight headless browser for AI agents.
Single Rust binary · 4 language SDKs · 33 MCP tools. Navigate URLs, extract structured content (markdown, links, screenshots), and build autonomous agent workflows.
📖 Full documentation: https://B4N1-com.github.io/b4n1-web/book/
🌍 Languages / Idiomas / 语言
| 🇬🇧 English | 🇪🇸 Español | 🇫🇷 Français | 🇩🇪 Deutsch | 🇵🇹 Português | 🇮🇹 Italiano |
| 🇨🇳 简体中文 | 🇯🇵 日本語 | 🇰🇷 한국어 | 🇷🇺 Русский | 🇸🇦 العربية | 🇮🇳 हिन्दी |
🖥 Platform Support
8 pre-compiled binaries — works everywhere:
| Platform | Architectures | Binary |
|---|---|---|
| Linux | x86_64, aarch64, i686 | musl (static, no glibc) |
| macOS | x86_64, arm64 | universal |
| Windows | x86_64, arm64 | MSVC |
Quick Start
# Install the binary
curl -sL https://raw.githubusercontent.com/B4N1-com/b4n1-web/master/scripts/install.sh | bash
# Or use your preferred package manager
pip install b4n1-web
npm install b4n1-web
dotnet add package B4n1Web
# Java: add dependency from Maven Central
from b4n1web import AgentBrowser
browser = AgentBrowser()
page = browser.goto("https://example.com")
print(page.markdown)
browser.close()
MCP Server
# stdio mode (default)
b4n1web mcp
npx b4n1-web mcp
uvx b4n1-web mcp
SDK Matrix
| Language | Package | Version | Binary |
|---|---|---|---|
| Python | b4n1-web | 0.16.0 | Bundled (musl) |
| JavaScript/TypeScript | b4n1-web | 0.16.0 | Bundled (musl) |
| Java | com.b4n1:b4n1-web | 0.16.0 | Bundled (musl) |
| C# (.NET) | B4n1Web | 0.16.0 | Bundled (musl) |
| Ruby | b4n1-web | 0.16.0 | Resolver (env/system/PATH) |
| PHP | b4n1/b4n1-web | 0.16.0 | Resolver (env/system/PATH) |
| Go | b4n1-web/sdks/go | 0.16.0 | Resolver (env/system/PATH) |
Capability parity
Every method each SDK exposes is tested (see coverage). Surface differs by
design: Python additionally ships a full MCP client (b4n1web.mcp), so it
covers the whole tool surface; JS/C#/Java/Ruby/PHP/Go wrap the CLI binary
(core flows). Cells below are measured against each SDK's actual source
(b4n1-web 0.16.0), not aspirational: explicit rounding — ✓ implemented ·
· not implemented · mcp available through the bundled MCP server.
| Capability | Python | JS/TS | C# | Java | Ruby | PHP | Go |
|---|---|---|---|---|---|---|---|
| goto (light / js / render) + page screenshot | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
wait (--wait-for / waitForSelector) | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
| click / typeText | ✓ | ✓ | ✓ | ✓ | · | · | · |
| links (Page) + find-links-by-text | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
| getLinks (browser) / one-shot fetch | mcp | ✓ | ✓ | · | · | · | · |
| evaluate (JS execution) | ✓ | · | · | · | ✓ | ✓ | ✓ |
| viewport / user-agent / emulate-device | ✓ | ✓ | · | · | · | · | · |
| upload / download / frames / cookies | ✓ | · | · | · | · | · | · |
| binary version query (bundle == SDK) | ✓ | ✓ | ✓ | ✓ | · | · | · |
Full parity = one MCP client per SDK (JS/CS/Java) — tracked as tasks; the Ruby/PHP/Go interaction gaps (click/type, device emulation, file transfer, cookies/frames) are their own parity tasks.
Browser Modes
| Mode | Description | RAM | Startup |
|---|---|---|---|
| Light | HTTP fetch + HTML parsing | ~15MB | Instant |
| JS | Light + JavaScript extraction | ~15MB | Instant |
| Render | Full Chromium + screenshots | ~100MB | ~2s |
Features
- 35 MCP tools for AI agent integration
- 4 language SDKs (Python, JS, Java, C#) with bundled binary
- Self-contained binary ~12MB, no dependencies
- Static linking (musl) — works on any Linux, no glibc required
- Three modes: Light (instant), JS (scripts), Render (Chromium)
- Security shield: domain filtering, safe browsing
- Network interception: block resources, mock responses
- MCP Server: stdio transport, no port needed
- Link extraction: get_links MCP tools + getLinks() in every SDK (one-shot fetch without a browser instance)
- --licenses flag: prints the bundled license + third-party notices, no key required
Documentation
- 📖 Full documentation — mdBook
- Agent Reference — AI agent integration
- MCP Tools — MCP registry
Links
- Website: https://b4n1.com
- GitHub: https://github.com/B4N1-com/b4n1-web
- PyPI: https://pypi.org/project/b4n1-web
- npm: https://www.npmjs.com/package/b4n1-web
- NuGet: https://www.nuget.org/packages/B4n1Web
- Maven Central: https://central.sonatype.com/artifact/com.b4n1/b4n1-web
License & Commercial Terms
- Original Source Code: 100% Proprietary and Closed Source (All Rights Reserved by Bani Montoya).
- Binaries & SDK Tools: Provided under the Business Source License 1.1 (BUSL-1.1).
- Free Tier: 100% free for development, evaluation, testing, personal projects, and startups with gross annual revenue < $100,000 USD.
- Mandatory Enterprise B2B License: Required for government agencies, public bidding projects, and enterprises with gross annual revenue >= $100,000 USD.
For Enterprise B2B licensing inquiries, visit https://b4n1.com/licensing or email [email protected].
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.