xarticle-mcp
Enables saving X (Twitter) Articles as Obsidian-faithful Markdown files with locally downloaded images and videos.
README
xarticle-mcp
Save an X (Twitter) Article to disk as Obsidian-faithful Markdown — with images and videos downloaded locally — so it reads in Obsidian the way it reads on X.
One tool: xarticle <url> → creates <slug>/<slug>.md + images/ (+ media/ for video) in your working directory, with YAML frontmatter and local, rewritten links.
Works in any MCP client: Claude Code, Codex, Cursor, Windsurf, … It fetches with your own X session, stored encrypted on your machine.
Requirements
- Node.js ≥ 18
- Google Chrome or Microsoft Edge installed (Edge ships with Windows). The tool drives your existing browser — no separate browser download.
Install (one line)
Add this to your MCP client config:
{
"mcpServers": {
"xarticle": { "command": "npx", "args": ["-y", "xarticle-mcp"] }
}
}
- Claude Code —
.mcp.jsonin your project root (or your user MCP config). - Codex — add an MCP server entry with the same
command+args. - Cursor / Windsurf / others — same
command+argsin their MCP settings.
Prefer running straight from source? Use
"args": ["-y", "github:aniirude/xarticle-mcp"].
macOS/Linux use the same config. The tool uses your installed Chrome/Edge; nothing else to install.
One-time login (paste 2 cookies)
X Articles need your logged-in session. There's no automated login (X rate-limits those), so you paste two session cookies once — they're stored encrypted at ~/.xarticle/.
npx -y xarticle-mcp login
It walks you through it:
- In a browser logged into X, press F12.
- Application tab (Chrome/Edge) or Storage (Firefox) → Cookies →
https://x.com. - Copy the value of
auth_token, paste, Enter. - Copy the value of
ct0, paste, Enter.
Check it anytime with npx -y xarticle-mcp status.
Alternative (no DevTools): npx -y xarticle-mcp login --browser reuses your real Chrome profile — but you must fully quit Chrome first (it can be flaky on Windows).
Use
In your MCP client, ask:
xarticle https://x.com/<user>/status/<id>
It writes, in your current working directory:
<article-slug>/
<article-slug>.md # frontmatter + body, local links
images/ # 01.jpg, 02.jpg, … (+ cover)
media/ # <slug>-01.mp4, … (videos, when downloadable)
- Saves to the directory your MCP client is running in. Pass
outputDirto override. - Videos with a direct MP4 are embedded as
![[…mp4]](Obsidian renders a player). Stream-only (HLS) videos fall back to a poster image + a link to X.
Tool input
| field | type | default | notes |
|---|---|---|---|
url |
string | — | the X Article/post URL |
outputDir |
string | working dir | where to create the folder |
imageFormat |
"original" | "png" |
original |
png re-encodes (needs sharp) |
CLI (handy for testing)
npx -y xarticle-mcp save <url> [outputDir] # fetch without an MCP client
npx -y xarticle-mcp status # check the saved session
Notes & caveats
- Personal use. This automates access with your own session; keep it low-volume. Respect X's Terms.
- Only what you can see. Articles your account can't view won't fetch.
- Obsidian video embeds (
![[…mp4]]) render inside an Obsidian vault. - X changes its markup. All X-DOM selectors live in
src/fetchArticle.ts— the one file to update if extraction drifts.
Develop
npm install
npm run build
npm run smoke # offline: markdown/video conversion + tools/list
MIT © aniirude
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。