My-Obscura-mcp
My-Obscura-mcp
README
<h2 align="center">Obscura</h2>
<p align="center"> <strong>Rust-native headless browser with built-in MCP server — no Chrome required</strong> </p>
<p align="center"> <img src="https://img.shields.io/badge/rust-1.75%2B-orange" alt="Rust 1.75+" /> <img src="https://img.shields.io/badge/MCP-stdio-blue" alt="MCP stdio" /> <img src="https://img.shields.io/badge/version-0.3.0-green" alt="v0.3.0" /> <img src="https://img.shields.io/badge/license-Apache--2.0-lightgrey" alt="Apache-2.0" /> </p>
What is Obscura?
Obscura is a Rust-based headless browser engine that provides both a Chrome DevTools Protocol (CDP) server and an MCP (Model Context Protocol) server. Run web automation, scraping, and AI-agent workflows without Chrome, Chromium, or Node.js — just a single native binary (~40MB) with built-in stealth mode and AI-optimized content extraction.
Why Obscura?
| Feature | Obscura | puppeteer-mcp | playwright-mcp |
|---|---|---|---|
| Chrome required | ❌ No | ✅ Yes | ✅ Yes |
| Binary size | ~40MB | N/A (needs Node+Chrome) | N/A (needs Node+Chromium) |
| Stealth mode | ✅ Built-in | ❌ No | ❌ No |
| AI-native extraction | ✅ LP domain | ❌ No | ❌ No |
| Language | Rust | Node.js | Node.js |
| Dep. services | ❌ None | Node.js ecosystem | Node.js ecosystem |
Quick Start
Download prebuilt binary (recommended)
Linux x86_64:
curl -fsSL https://github.com/SmongsDev/my-obscura-mcp/releases/latest/download/obscura-x86_64-unknown-linux-gnu.tar.gz | tar xz
chmod +x obscura && sudo mv obscura /usr/local/bin/
obscura --version
Linux ARM64:
curl -fsSL https://github.com/SmongsDev/my-obscura-mcp/releases/latest/download/obscura-aarch64-unknown-linux-gnu.tar.gz | tar xz
chmod +x obscura && sudo mv obscura /usr/local/bin/
obscura --version
macOS (Apple Silicon):
curl -fsSL https://github.com/SmongsDev/my-obscura-mcp/releases/latest/download/obscura-aarch64-apple-darwin.tar.gz | tar xz
chmod +x obscura && sudo mv obscura /usr/local/bin/
obscura --version
macOS (Intel):
curl -fsSL https://github.com/SmongsDev/my-obscura-mcp/releases/latest/download/obscura-x86_64-apple-darwin.tar.gz | tar xz
chmod +x obscura && sudo mv obscura /usr/local/bin/
obscura --version
Windows (PowerShell):
$ProgressPreference = 'SilentlyContinue'
Invoke-WebRequest -Uri "https://github.com/SmongsDev/my-obscura-mcp/releases/latest/download/obscura-x86_64-pc-windows-msvc.zip" -OutFile obscura.zip
Expand-Archive obscura.zip -DestinationPath .
.\obscura.exe --version
Build from Source
Prerequisites: Rust 1.75+ (install via rustup.rs)
cargo build --release
Note: First build takes ~5 minutes due to V8 compilation. Subsequent builds use cached artifacts.
The binary will be at target/release/obscura.
MCP Setup
Claude Desktop (macOS / Linux / Windows)
Edit ~/.config/Claude/claude_desktop_config.json (or %APPDATA%\Claude\claude_desktop_config.json on Windows):
{
"mcpServers": {
"obscura": {
"command": "~/.cargo/bin/obscura",
"args": ["mcp"]
}
}
}
Note: Replace ~/.cargo/bin/obscura with the full path to your binary if it's not in the default Cargo installation directory.
Restart Claude Desktop. Obscura tools will appear in the Tools panel.
Cursor
Edit .cursor/mcp.json:
{
"mcpServers": {
"obscura": {
"command": "~/.cargo/bin/obscura",
"args": ["mcp"]
}
}
}
Restart Cursor.
Claude Code (CLI)
Add to your Claude Code settings:
# ~/.claude/settings.json or ~/.claude/settings.local.json
{
"mcpServers": {
"obscura": {
"command": "~/.cargo/bin/obscura",
"args": ["mcp"]
}
}
}
Then restart the Claude Code CLI.
Available Tools
All 29 tools (22 existing + 7 new in v0.3.0):
| Tool | Description |
|---|---|
browser_navigate |
Navigate to a URL. Call this first before any extraction. |
browser_evaluate |
Evaluate a JavaScript expression on the current page. |
browser_get_page_status |
Fetch current URL, page title, and lifecycle state (Idle/Loading/Loaded/NetworkIdle). |
browser_get_html |
Get the full page source HTML. |
browser_get_markdown |
Convert page to Markdown format. |
browser_get_readable_content |
Extract main article text, removing ads and navigation. Returns text + word count. |
browser_get_metadata |
Extract meta tags: title, description, OG (Open Graph), Twitter Card, canonical URL, etc. |
browser_get_links |
Get all links on page as array of { href, text }. All URLs converted to absolute. |
browser_get_forms |
Extract form structure: input types, labels, select options, field names. |
browser_get_table_data |
Convert HTML tables to structured JSON with header-based keys. |
browser_get_structured_data |
Extract JSON-LD, OpenGraph, and microdata schemas. |
browser_get_element_text |
Get text content of an element by CSS selector. |
browser_get_element_attribute |
Read HTML attribute value (href, src, data-*, etc.) by CSS selector. |
browser_click |
Click an element by CSS selector. Triggers click events. |
browser_type_text |
Type text into an input field. Triggers input/change events. |
browser_select |
Select an option in a <select> dropdown by value or visible text. |
browser_get_elements |
Query multiple elements by CSS selector, return array of { text, attributes }. |
browser_get_headings |
Extract all headings (h1–h6) as array of { level, text }. |
browser_get_all_images |
Get all images with { src, alt, title, width, height }. |
browser_find_text |
Find elements containing keyword text. Returns tag, id, class, and text (max 50 matches). |
browser_scroll |
Scroll to coordinates or element into view. Useful for lazy-loaded content. |
browser_keyboard_press |
Dispatch keyboard events (Enter, Escape, etc.) to focused element or target by selector. |
browser_navigate_and_extract |
Combine navigate + readable_content in one call (reduces round-trips). |
browser_get_console_logs |
Capture all console.log/warn/error output from the page. |
browser_set_emulation |
Spoof User-Agent, viewport dimensions. Use for mobile page access. |
browser_wait_for_selector * |
Wait for an element to appear in the DOM (useful for async rendering). |
browser_get_cookies * |
Retrieve all cookies for the current page. |
browser_set_cookie * |
Set a cookie (name, value, domain, path, expiry, etc.). |
browser_delete_cookie * |
Delete a cookie by name. |
browser_get_local_storage * |
Fetch all localStorage items for the current origin. |
browser_set_local_storage * |
Set a localStorage key–value pair. |
browser_clear_storage * |
Clear localStorage and sessionStorage. |
* = New in v0.3.0
Stealth Mode
Obscura includes built-in stealth mode to avoid bot detection. Use the --stealth flag when starting:
obscura mcp --stealth
Or set via CDP/MCP requests. Stealth mode spoofs TLS fingerprints and browser identifiers.
License
Apache 2.0 — see LICENSE
<details> <summary>한국어</summary>
개요
Obscura는 Chrome 없이 웹 자동화, 스크래핑, AI 에이전트 워크플로우를 실행할 수 있는 Rust 기반 헤드리스 브라우저 엔진입니다. CDP(Chrome DevTools Protocol) 서버와 MCP(Model Context Protocol) 서버를 제공하며, 단일 네이티브 바이너리(~40MB)로 구동되며 내장 스텔스 모드와 AI 최적화 콘텐츠 추출 기능을 갖추고 있습니다.
제공 도구
모든 29개 도구 (기존 22개 + v0.3.0 신규 7개):
| 도구 | 설명 |
|---|---|
browser_navigate |
URL 탐색. 다른 도구 사용 전 반드시 먼저 호출 |
browser_evaluate |
현재 페이지에서 JS 표현식 평가 |
browser_get_page_status |
현재 URL / 제목 / 라이프사이클 상태 |
browser_get_html |
페이지 원본 HTML |
browser_get_markdown |
페이지 → Markdown 변환 |
browser_get_readable_content |
광고·네비게이션 제거 후 본문 텍스트 + 단어 수 |
browser_get_metadata |
제목, 설명, OG 태그, Twitter Card, 카노니컬 URL 등 |
browser_get_links |
페이지 내 전체 링크 (절대 URL 변환 포함) |
browser_get_forms |
폼 필드 구조 (입력 타입, 레이블, 셀렉트 옵션 등) |
browser_get_table_data |
표 → 헤더 키 기반 구조화 JSON |
browser_get_structured_data |
JSON-LD / OpenGraph / microdata 추출 |
browser_get_element_text |
CSS 셀렉터로 요소 텍스트 추출 |
browser_get_element_attribute |
CSS 셀렉터로 속성값 읽기 (href, src, data-* 등) |
browser_click |
CSS 셀렉터로 요소 클릭 |
browser_type_text |
입력 필드에 텍스트 입력 (input/change 이벤트 발생) |
browser_select |
<select> 드롭다운 옵션 선택 (값 또는 텍스트로 매칭) |
browser_get_elements |
querySelectorAll로 여러 요소의 텍스트·속성 배열 반환 |
browser_get_headings |
h1~h6 전체를 [{ level, text }] JSON으로 반환 |
browser_get_all_images |
img 태그의 src, alt, title, width, height 목록 |
browser_find_text |
키워드 포함 요소 텍스트·태그·id·클래스명 반환 (최대 50건) |
browser_scroll |
좌표 스크롤 또는 요소 scrollIntoView (lazy load 트리거용) |
browser_keyboard_press |
CSS 셀렉터 요소 또는 포커스 요소에 키보드 이벤트 디스패치 |
browser_navigate_and_extract |
navigate + readable_content 단일 호출로 라운드트립 절감 |
browser_get_console_logs |
페이지 JS console.* 캡처 로그 배열 반환 |
browser_set_emulation |
User-Agent·뷰포트 너비/높이 설정 (모바일 페이지 접근용) |
browser_wait_for_selector * |
DOM에 요소가 나타날 때까지 대기 (비동기 렌더링용) |
browser_get_cookies * |
현재 페이지의 모든 쿠키 조회 |
browser_set_cookie * |
쿠키 설정 (이름, 값, 도메인, 경로, 만료 등) |
browser_delete_cookie * |
쿠키 삭제 |
browser_get_local_storage * |
현재 origin의 localStorage 전체 조회 |
browser_set_local_storage * |
localStorage 키–값 쌍 설정 |
browser_clear_storage * |
localStorage·sessionStorage 전체 삭제 |
* = v0.3.0 신규
스택
- 언어: Rust 1.75+ · Tokio
current_thread - JS 엔진: Deno Core / V8
- HTML 파싱: html5ever
- MCP 프레임워크: rmcp
라이선스
Apache 2.0 — LICENSE 참고
</details>
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。