Why this matters right now
Platforms like Xiaohongshu are where creators publish, test, and iterate content rapidly — but most LLM-based assistants lack standardized, first-class tools to operate those workflows. xiaohongshu-mcp bridges that gap by exposing Xiaohongshu actions as MCP tools, turning content publishing, search, and interaction into callable operations for AI agents and pipelines.
What Sets It Apart
- Exposes a compact set of MCP tools (check_login_status, get_login_qrcode, publish_content, publish_with_video, list_feeds, search_feeds, get_feed_detail, post_comment_to_feed, etc.), so an LLM or MCP-capable client can script publishing and data retrieval without page-scraping logic.
- Browser-automation backend with headless / non-headless modes and Docker images — you can run it locally or in containerized environments and keep credentials off remote servers; HTTP MCP runs on localhost:18060 by default for easy client integration.
- Designed for MCP-first clients (Claude Code, Cursor, VSCode integrations described), plus a lightweight browser-extension alternative (x-mcp) for non-technical users who want zero-deploy setup.
- Community-oriented: multi-contributor repo with extensive README, usage examples, and ecosystem tooling (OpenClaw/OpenClaw skill integrations, third-party wrappers), so it’s ready for practical operator workflows rather than experimental PoC.
Who It's For and Tradeoffs
Great fit if you are building LLM-driven creator tools, automated content pipelines, or AI agents that must publish/search/interact on Xiaohongshu and you want a standardized MCP interface rather than bespoke scraping code. It’s also suitable for teams who prefer running automation locally or in controlled Docker instances.
Look elsewhere if you need a cloud-hosted, fully managed posting service (this project expects local execution and manages login cookies/captured credentials) or if you cannot accept the platform-level risks: automated posting and scripted interactions can trigger platform anti-bot measures, violate Xiaohongshu terms of service in some cases, or require manual intervention for account verification. Operational considerations: you must provide valid login credentials/cookies, handle rate limits and captchas, and maintain the browser automation dependencies (Chrome/Chromedriver or equivalent).
Notes and provenance
- Repository owner: xpzouying (GitHub). First public activity and wider community attention date back to August 2025 (project evolved through Aug–Nov 2025 and continued maintenance into 2026). The repo implements the MCP server primarily in Go and ships README, Dockerfiles, and a browser-plugin variant for simpler usage.
