xpzouying/xiaohongshu-mcp

xiaohongshu‑mcp

What it is – A self‑hosted service that exposes the Xiaohongshu (小红书) social‑media platform as a Model Context Protocol (MCP) server. By running the server, an AI assistant (Claude Code, Cursor, VS Code, Open Code, Cline, etc.) can call Xiaohongshu functions such as login, posting, searching, liking, commenting, and fetching user or feed details.

Key capabilities

Login management – QR‑code login, cookie handling, status check, cookie reset.
Content publishing – Post image‑text notes (title, body, images) or video notes (title, body, local video). Supports tags, scheduling, visibility options, and product linking.
Feed operations – Retrieve the home recommendation list, search by keyword with filters, get full post details (including comments and interaction stats), like/unlike, favorite/unfavorite.
Social interactions – Post comments, reply to specific comments, fetch a user’s profile and their notes.
MCP integration – All functions are exposed as standard MCP tools (check_login_status, publish_content, search_feeds, …) that any MCP‑compatible client can call.

How to get it running

Pre‑built binaries – Download the appropriate executable for macOS Apple Silicon, Windows x64, or Linux x64 from the Releases page. Run the xiaohongshu-login-* binary once to log in, then start xiaohongshu-mcp-*.
Docker – Pull the image xpzouying/xiaohongshu-mcp and launch it with the provided docker-compose.yml. The container automatically installs a headless Chromium browser, mounts ./data for cookies and ./images for uploaded media, and exposes port 18060.
Build from source – Requires Go 1.20+. Set a GOPROXY (e.g., https://goproxy.cn) and run go build ./....

Typical usage flow

# 1. Log in (once per session)
./xiaohongshu-login-darwin-arm64   # or the Windows/Linux binary

# 2. Start the MCP server (default headless mode)
./xiaohongshu-mcp-darwin-arm64
#   → listens on http://localhost:18060/mcp

# 3. Verify with the MCP inspector
npx @modelcontextprotocol/inspector
#   → connect to the above URL and list available tools

Connecting an AI client – Add the server URL (http://localhost:18060/mcp) to the client’s MCP configuration. Examples are provided for:

Claude Code CLI (claude mcp add …)
Open Code CLI (interactive opencode mcp add)
Cursor (.cursor/mcp.json)
VS Code (.vscode/mcp.json)
Google Gemini CLI, Cline, and any generic HTTP‑MCP client.

Security & optional auth

By default the server is open on localhost. For production use set AUTH_TOKEN (or the -token flag) and have clients send Authorization: Bearer <token>.
Proxy support via XHS_PROXY (HTTP/HTTPS/SOCKS5) for environments that need outbound routing.

Community & support

29 contributors (All‑Contributors badge).
Docker pull stats, star‑history chart, and a dedicated donations page (all proceeds go to charity).
Extensive troubleshooting guide (Issue #56) and a separate browser‑extension version (xpzouying/x-mcp) for non‑technical users.

Who might use this

Developers building AI agents that need to read or publish Xiaohongshu content.
Content creators who want to automate posting, commenting, or data collection via prompts.
Researchers experimenting with social‑media‑driven LLM workflows.

All details above are taken directly from the repository’s README; no additional features are inferred.

xpzouying/xiaohongshu-mcp

xiaohongshu‑mcp

Related