Xiaohongshu MCP: Automate RedNote's Most Viral Platform from Your Terminal
15,341 downloads, 59 stars, 118 installs. The Xiaohongshu MCP skill by @Borye automates content operations on Xiaohongshu (小红书), the platform known internationally as RedNote. With 300 million monthly active users, Xiaohongshu is China's dominant lifestyle and product discovery platform — part Instagram, part Pinterest, part user review site. This skill wraps the xpzouying/xiaohongshu-mcp server (8,400+ GitHub stars) into a Python CLI that Claude can drive directly.
Why This Platform Matters Right Now
Xiaohongshu had its biggest international moment in early 2025, when fears of a US TikTok ban drove millions of American users to the platform (branding it "RedNote"). The numbers tell the story: US downloads jumped from 509,000 in Q4 2024 to 3.7 million in Q1 2025 — a 627% increase. Daily active US users peaked at 1.3 million in January 2025, and by March 2025 had stabilized at 800,000 — still 114% above December 2024 levels. The spike was temporary, but the international awareness it created was not.
For brands operating in or expanding into China — or for any business that wants to reach Chinese consumers internationally — Xiaohongshu is a mandatory channel. The platform is particularly dominant in fashion, beauty, food, travel, and consumer electronics. A note that goes viral on Xiaohongshu can drive more product discovery than most other platforms combined for certain Chinese demographics.
Until recently, programmatic access required building custom scrapers or navigating unofficial API reverse-engineering. This skill provides a clean, maintained, QR-auth-based interface backed by a server with strong community validation (8,400+ stars).
Architecture
The skill requires two components to run together:
Your Agent (Claude + Skill)
│
▼
Python Client (xhs_client.py)
│ HTTP requests to localhost:18060
▼
xiaohongshu-mcp Server (local binary)
│ Automated browser session
▼
Xiaohongshu API
The local MCP server runs a managed browser session authenticated to your Xiaohongshu account via QR code login. The Python client communicates with it over HTTP. Claude drives the Python client via CLI commands.
Setup
Step 1: Download the MCP Server Binary
# macOS (Apple Silicon)
wget https://github.com/xpzouying/xiaohongshu-mcp/releases/latest/download/xiaohongshu-mcp-darwin-arm64
wget https://github.com/xpzouying/xiaohongshu-mcp/releases/latest/download/xiaohongshu-login-darwin-arm64
chmod +x xiaohongshu-mcp-darwin-arm64 xiaohongshu-login-darwin-arm64
# Windows: download xiaohongshu-mcp-windows-amd64.exe
# Linux: download xiaohongshu-mcp-linux-amd64Step 2: Login (First Time Only)
./xiaohongshu-login-darwin-arm64
# Opens browser with QR code — scan with your Xiaohongshu mobile appImportant: Don't log into the same Xiaohongshu account in any other browser while the server is running — this invalidates the session.
Step 3: Start the Server
./xiaohongshu-mcp-darwin-arm64 # headless (recommended)
./xiaohongshu-mcp-darwin-arm64 -headless=false # visible browser for debuggingThe server runs at http://localhost:18060.
Step 4: Install the Skill
clawhub install xiaohongshu-mcpCore Commands
Check Login Status
python scripts/xhs_client.py status
# ✅ Logged in as: your_username
# ❌ Not logged in. Please run the login tool first.Search Notes
# Basic search
python scripts/xhs_client.py search "户外电源" # "portable power station"
# With filters
python scripts/xhs_client.py search "咖啡推荐" \
--sort "最多点赞" \ # sort by most liked
--type "图文" \ # image+text notes only
--time "一周内" # within last week
# Output includes feed_id and xsec_token for each resultGet Note Details and Comments
After searching, use the feed_id and xsec_token from search results:
python scripts/xhs_client.py detail "note_feed_id" "xsec_token_here"
# Returns: full note text, author, location, engagement stats
# + top comments with author namesThe two-step search → detail workflow mirrors how the platform itself works and ensures you get accurate, real-time data.
Get Recommended Feed
python scripts/xhs_client.py feeds
# Returns: current trending/recommended notes on your feedUseful for trend monitoring — what's currently surfacing on the platform in a given category.
Publish a Note
python scripts/xhs_client.py publish \
"产品评测：这款便携电源真的很香" \ # title
"开箱体验分享，续航测试..." \ # content
"https://cdn.example.com/img1.jpg,https://cdn.example.com/img2.jpg" \
--tags "好物推荐,数码,便携电源"Publishes a new image+text note to your account. Images must be publicly accessible URLs.
Use Cases
Brand monitoring. Automatically search for mentions of your brand or product category on Xiaohongshu, retrieve full note content and comment sentiment, and generate daily monitoring reports — all without manual browsing.
Market research. Before launching in a new category, search top-performing notes in that space, analyze what drives engagement (likes, saves, comments), and extract consumer language and pain points from comments.
Content repurposing. For content teams that create notes manually, Claude can assist with drafting titles and descriptions optimized for Xiaohongshu's visual-first format, then publish via the skill.
Trend detection. Monitor the recommended feed daily, track emerging topics before they peak, and surface early signals for product or content teams.
Competitive intelligence. Systematically pull notes from competitor brands, track their posting frequency and engagement rates, and identify which content formats perform best in your category.
Practical Tips
The two-token requirement is intentional. Xiaohongshu note detail requests require both a feed_id and an xsec_token (a security token generated per search result). Always run search first, then use the tokens from those results for detail requests.
Headless mode is more stable. The -headless=false flag is for debugging only. In production, always run headless — the visible browser window can cause session instability on some systems.
Session persistence matters. If your workflow runs for many hours, the session may time out. Build a status check (xhs_client.py status) at the start of each workflow run to verify the server is logged in before proceeding.
Sort options are Chinese-language flags. The --sort and --type and --time filters use Chinese strings (综合, 最新, 最多点赞, etc.). Copy them directly from the documentation rather than translating.
Considerations
- Local server required. Unlike API-key-based skills, this one requires a running local server process. It cannot be deployed as a serverless or remote function without additional infrastructure. The server must be running on the same machine as the agent.
- Session invalidation risk. If you or anyone else logs into the same Xiaohongshu account in a browser while the server is running, the session breaks. Dedicate a separate account for automation.
- Publishing limits. Xiaohongshu has daily publishing limits and rate limits for new accounts. Automation-heavy posting on a fresh account may trigger risk controls. Use an established account.
- Platform ToS. Automated interactions with Xiaohongshu exist in a gray area relative to the platform's terms of service. Review current terms before deploying at scale, particularly for commercial campaigns.
- China network access. Xiaohongshu is a Chinese platform. Running the server outside China may require a VPN or proxy if direct access is restricted in your region.
The Bigger Picture
Xiaohongshu sits at an interesting intersection: it's arguably the most influential product discovery platform in China, with growing international relevance, yet it remains largely inaccessible to the AI tooling ecosystem that has emerged around Western social media platforms.
The 8,400+ stars on the underlying xpzouying/xiaohongshu-mcp server reflect genuine demand. The Xiaohongshu MCP skill packages that project into a ClawHub-compatible format, making it accessible to the wider OpenClaw agent ecosystem.
For brands, marketers, and researchers who need programmatic access to Xiaohongshu data, this is currently one of the only reliable, maintained options available.
View the skill on ClawHub: xiaohongshu-mcp