xiaohongshu-skill
MCP server for automating Xiaohongshu (RED Note). Publish posts, search content, comment, like, and analyze creator data on China's largest lifestyle social platform.
README
<p align="center"> <h1 align="center">π xiaohongshu</h1> <p align="center">AI-powered Xiaohongshu (ε°ηΊ’δΉ¦/RED) automation β publish, search, engage, and analyze.</p> </p>
<p align="center"> <a href="#cli-quick-start">CLI Quick Start</a> β’ <a href="#features">Features</a> β’ <a href="#quick-start">Quick Start</a> β’ <a href="#mcp-tools">MCP Tools</a> β’ <a href="#cdp-scripts">CDP Scripts</a> β’ <a href="#openclaw-integration">OpenClaw</a> β’ <a href="#claude-code-integration">Claude Code</a> β’ <a href="./LICENSE">License</a> </p>
<p align="center"> <img src="./demo.gif" alt="xiaohongshu demo" width="800"> </p>
CLI Quick Start
The fastest way to get started β 3 commands from zero to posting:
# 1. Clone
git clone https://github.com/Youhai020616/xiaohongshu.git
cd xiaohongshu
# 2. One-click install (auto: Python check β venv β pip install β MCP binary)
bash setup.sh
# 3. Initialize (auto: proxy config β start MCP β QR login)
source activate.sh && xhs init
Then just use:
xhs search "AIεδΈ" # Search
xhs publish -t "Hello" -c "My first post" -i photo.jpg # Publish
xhs like FEED_ID -t TOKEN # Like
xhs analytics # Dashboard
xhs --help # All commands
π Full CLI guide: docs/cli-guide.md
What is this?
A complete toolkit for automating Xiaohongshu (ε°ηΊ’δΉ¦/RED Note) operations through two complementary engines:
| Engine | Technology | Use Cases | Startup |
|---|---|---|---|
| MCP Server | Go binary, JSON-RPC | Publish, search, comment, like, follow | Always-on daemon |
| CDP Scripts | Python, Chrome DevTools | Analytics dashboard, notifications, advanced search | On-demand |
Built as an OpenClaw Skill, but works standalone or with any MCP-compatible client (Claude Code, Cursor, etc.).
Features
- π Publish β Image posts and video posts with tags, scheduling, and visibility control
- π Search β Keyword search with filters (sort, note type, suggested keywords)
- π¬ Engage β Comment, reply, like, favorite on any post
- π€ Profile β Fetch any user's profile and posts
- π Analytics β Creator dashboard data export (CSV), content performance metrics
- π Notifications β Fetch mentions and interaction notifications
- π₯ Multi-Account β Isolated Chrome profiles per account
- π QR Code Login β Scan-to-login, persistent cookie storage
Quick Start
Prerequisites
- macOS (Apple Silicon) β MCP binary is pre-built for
darwin-arm64 - Python 3.9+ (for CDP scripts)
- Google Chrome (for CDP scripts and QR login)
- A proxy/VPN if outside China (MCP requires
-rodproxy flag)
1. Clone
git clone https://github.com/Youhai020616/xiaohongshu.git
cd xiaohongshu
2. Start MCP Server
cd mcp
chmod +x xiaohongshu-mcp-darwin-arm64
# Without proxy (inside China)
./xiaohongshu-mcp-darwin-arm64 -port :18060
# With proxy (outside China)
./xiaohongshu-mcp-darwin-arm64 -port :18060 -rod "proxy=http://127.0.0.1:7897"
3. Login
On first run, get a QR code to scan with your Xiaohongshu app:
curl -s -X POST http://127.0.0.1:18060/mcp \
-H "Content-Type: application/json" \
-H "Accept: application/json, text/event-stream" \
-d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2024-11-05","capabilities":{},"clientInfo":{"name":"cli","version":"1.0"}}}'
# Save the Mcp-Session-Id from response headers
curl -s -X POST http://127.0.0.1:18060/mcp \
-H "Content-Type: application/json" \
-H "Accept: application/json, text/event-stream" \
-H "Mcp-Session-Id: YOUR_SESSION_ID" \
-d '{"jsonrpc":"2.0","method":"notifications/initialized"}'
curl -s -X POST http://127.0.0.1:18060/mcp \
-H "Content-Type: application/json" \
-H "Accept: application/json, text/event-stream" \
-H "Mcp-Session-Id: YOUR_SESSION_ID" \
-d '{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"get_login_qrcode","arguments":{}}}'
Or use the login helper binary:
cd mcp
./xiaohongshu-login-darwin-arm64
4. Publish Your First Post
curl -s -X POST http://127.0.0.1:18060/mcp \
-H "Content-Type: application/json" \
-H "Accept: application/json, text/event-stream" \
-H "Mcp-Session-Id: YOUR_SESSION_ID" \
-d '{
"jsonrpc":"2.0","id":3,
"method":"tools/call",
"params":{
"name":"publish_content",
"arguments":{
"title":"Hello from API π",
"content":"My first automated post on Xiaohongshu!",
"images":["/path/to/image.jpg"],
"tags":["ζ΅θ―","θͺε¨ε"]
}
}
}'
MCP Tools
The MCP server exposes these tools via the Model Context Protocol:
| Tool | Description | Key Parameters |
|---|---|---|
check_login_status |
Check if logged in | β |
get_login_qrcode |
Get QR code for login (base64) | β |
delete_cookies |
Reset login state | β |
publish_content |
Publish image post | title, content, images, tags?, visibility?, is_original?, schedule_at? |
publish_with_video |
Publish video post | title, content, video (local path), tags?, visibility?, schedule_at? |
search_feeds |
Search posts by keyword | keyword, filters? |
get_feed_detail |
Get post details + comments | feed_id, xsec_token, load_all_comments? |
post_comment_to_feed |
Comment on a post | feed_id, xsec_token, content |
reply_comment_in_feed |
Reply to a comment | feed_id, xsec_token, comment_id, user_id, content |
like_feed |
Like / unlike a post | feed_id, xsec_token, unlike? |
favorite_feed |
Favorite / unfavorite a post | feed_id, xsec_token, unfavorite? |
list_feeds |
Get homepage feed | β |
user_profile |
Get user profile + posts | user_id, xsec_token |
get_self_info |
Get own account info | β |
MCP Session Protocol
The MCP server uses Streamable HTTP transport. Every request must include:
Content-Type: application/json
Accept: application/json, text/event-stream
Mcp-Session-Id: <from initialize response header>
Session lifecycle: initialize β notifications/initialized β tools/call (repeat)
Parameters Reference
publish_content
title(string, required) β Max 20 Chinese characterscontent(string, required) β Post body, max 1000 characters. Do NOT include hashtags hereimages(string[], required) β Local file paths or HTTP URLs (at least 1)tags(string[], optional) β Hashtags, e.g.["ζ θ‘", "ηΎι£"]visibility(string, optional) βε ¬εΌε―θ§(default) /δ» θͺε·±ε―θ§/δ» δΊε ³ε₯½εε―θ§is_original(bool, optional) β Declare as original contentschedule_at(string, optional) β ISO 8601 datetime, e.g.2026-03-15T10:30:00+08:00
publish_with_video
video(string, required) β Local absolute path to video file only
CDP Scripts
For features the MCP server doesn't support (analytics, notifications, suggested keywords), use the Python CDP scripts:
Setup
pip install -r requirements.txt
Chrome Launcher
# Start Chrome with isolated profile
python scripts/chrome_launcher.py
# Headless mode
python scripts/chrome_launcher.py --headless
# Stop Chrome
python scripts/chrome_launcher.py --kill
Available Commands
# Login (scan QR code)
python scripts/cdp_publish.py login
# Creator dashboard data β CSV
python scripts/cdp_publish.py content-data --csv-file output.csv
# Notification mentions
python scripts/cdp_publish.py get-notification-mentions
# Search with filters
python scripts/cdp_publish.py search-feeds --keyword "AIεδΈ" --sort-by ζζ° --note-type εΎζ
# Post detail
python scripts/cdp_publish.py get-feed-detail --feed-id FEED_ID --xsec-token TOKEN
# Comment
python scripts/cdp_publish.py post-comment-to-feed --feed-id FEED_ID --xsec-token TOKEN --content "Great post!"
# Publish via CDP (alternative to MCP)
python scripts/publish_pipeline.py --headless \
--title-file title.txt --content-file content.txt \
--image-urls "https://example.com/img.jpg"
Multi-Account
python scripts/cdp_publish.py add-account work --alias "Work Account"
python scripts/cdp_publish.py --account work login
python scripts/cdp_publish.py --account work content-data
python scripts/cdp_publish.py list-accounts
OpenClaw Integration
Install as an OpenClaw skill:
# Copy to skills directory
cp -r xiaohongshu ~/.openclaw/skills/xiaohongshu
# Start MCP server
cd ~/.openclaw/skills/xiaohongshu/mcp
./start.sh
The SKILL.md file provides full instructions for OpenClaw's AI agent to use the tools automatically.
Claude Code Integration
See docs/claude-code-integration.md for setup instructions with Claude Code.
Tips & Known Issues
- Login verification:
check_login_statushas a DOM detection bug β usesearch_feedswith any keyword to verify login state - Post ID:
publish_contentmay return an empty PostID β this is normal. Usesearch_feedswith your username to verify - Don't retry publishing β A timeout doesn't mean failure. Always verify before retrying
- Proxy required: Outside China, the MCP server needs
-rod "proxy=..."flag - Concurrent sessions: Don't log in to the same account from both MCP and web browser simultaneously
visibilityvalues: Must be in Chinese:ε ¬εΌε―θ§,δ» θͺε·±ε―θ§,δ» δΊε ³ε₯½εε―θ§schedule_atrange: Must be between 1 hour and 14 days from now
Platform Support
| Component | macOS ARM64 | macOS x86 | Linux | Windows |
|---|---|---|---|---|
| xhs CLI | β | β | β | β |
| MCP Server | β | β | β | β |
| CDP Scripts | β | β | β | β |
Windows / Linux / macOS x86 users: Run xhs init β the CLI auto-detects missing MCP binary and switches to CDP-only mode. All features (publish, search, analytics, etc.) work through Chrome.
# Windows
setup.bat # One-click install
activate.bat && xhs init # Initialize (auto CDP mode)
# macOS / Linux
bash setup.sh
source activate.sh && xhs init
Project Structure
xiaohongshu/
βββ README.md # This file
βββ SKILL.md # OpenClaw skill definition
βββ pyproject.toml # CLI package config (pip install -e .)
βββ manifest.json # Skill metadata
βββ LICENSE # MIT License
βββ requirements.txt # Python dependencies (legacy)
βββ .gitignore
βββ src/xhs_cli/ # β CLI package
β βββ main.py # Unified entry point (xhs command)
β βββ engines/
β β βββ mcp_client.py # MCP JSON-RPC client (auto session)
β β βββ cdp_client.py # CDP scripts wrapper
β βββ commands/
β β βββ init.py # xhs init (guided setup)
β β βββ auth.py # xhs login/logout/status
β β βββ publish.py # xhs publish (auto engine)
β β βββ search.py # xhs search/detail
β β βββ interact.py # xhs like/comment/favorite
β β βββ analytics.py # xhs analytics/notifications
β β βββ account.py # xhs account management
β β βββ profile.py # xhs me/profile
β β βββ server.py # xhs server start/stop/status
β β βββ config_cmd.py # xhs config show/set/get
β βββ utils/
β βββ config.py # ~/.xhs/config.json management
β βββ output.py # Rich formatted output
βββ mcp/
β βββ xiaohongshu-mcp-darwin-arm64 # MCP server binary
β βββ xiaohongshu-login-darwin-arm64 # Login helper binary
β βββ start.sh # Startup script
βββ scripts/
β βββ cdp_publish.py # Main CDP automation (2700 lines)
β βββ chrome_launcher.py # Chrome lifecycle management
β βββ publish_pipeline.py # High-level publish workflow
β βββ feed_explorer.py # Feed browsing utilities
β βββ account_manager.py # Multi-account management
β βββ image_downloader.py # Image download helper
β βββ run_lock.py # Process locking
βββ config/
β βββ accounts.json.example # Account config template
βββ docs/
βββ cli-guide.md # β CLI usage guide
βββ claude-code-integration.md # Claude Code setup guide
Contributing
Issues and PRs welcome! Areas where help is needed:
- [ ] Cross-platform MCP builds (Linux, Windows, macOS x86)
- [ ] MCP Go source code release
- [ ] Video upload improvements
- [ ] Rate limiting and retry logic
- [ ] Test suite
License
<p align="center"> <sub>Built with β€οΈ by <a href="https://github.com/Youhai020616">Youhai</a></sub> </p>
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.