apify-mcp-server vs xiaohongshu-mcp
Two of the top projects in MCP servers, side by side: who leads on each score part and who each one suits, then setup, license, activity and what each review found.
apify-mcp-server
Official MCP server exposing Apify Store Actors as tools for web scraping
xiaohongshu-mcp
MCP server that lets AI clients post, search and comment on Xiaohongshu
Verdict
apify-mcp-server ranks higher, 86 to 85. Of the 4 score parts we measure for both, apify-mcp-server leads on 2, xiaohongshu-mcp on 1 and 1 is even.
- Adoptionxiaohongshu-mcp leads by 6 points, 79 to 73.
- FreshnessEven at 100.
- Maintenanceapify-mcp-server leads by 1 point, 84 to 83: close to even.
- Agent-readyapify-mcp-server leads by 15 points, 85 to 70: a wide gap.
Pick apify-mcp-server if you are
Agent builders who need web scraping and data extraction tools over MCP
- One connection reaches thousands of Store Actors instead of a separate MCP server per site
- Tool categories selectable via `tools` URL parameter, `--tools` flag or `TOOLS` env var
Pick xiaohongshu-mcp if you are
Developers connecting an AI assistant to Xiaohongshu for posting and research
- Covers publishing, search, comments, likes, favorites and profiles as MCP tools
- Prebuilt binaries for macOS arm64, Windows x64 and Linux x64, plus Docker images
| What we compare | apify-mcp-server | xiaohongshu-mcp |
|---|---|---|
| Score parts, out of 100 | ||
| Adoption | 73, popular | 79, popular |
| Momentum | Not scored for this kind | Not scored for this kind |
| Freshness | 100, active | 100, active |
| Maintenance | 84, healthy | 83, healthy |
| Agent-ready | 85, ready | 70, partly |
| Facts from GitHub and the README | ||
| Stars | 15.7k | 17.1k |
| License | MIT (permissive) | Apache-2.0 (permissive) |
| Last commit | Nov 2026 | Nov 2026 |
| Last release | Nov 2026 | Nov 2026 |
| Language | TypeScript | Go |
| Docker | Yes | Yes |
| GPU | Not needed | Not needed |
| arm64 or Apple Silicon | Not stated | Mentioned |
apify-mcp-server
MCP server that lets an agent search Apify Store, read an Actor's input schema and pricing, run it, and fetch results from its dataset. It runs hosted at mcp.apify.com over Streamable HTTP with OAuth or a bearer token, or locally over stdio via the @apify/actors-mcp-server npm package. Actors execute on the Apify platform in both modes.
Who it is for: Agent builders who need web scraping and data extraction tools over MCP
Strengths
- One connection reaches thousands of Store Actors instead of a separate MCP server per site
- Tool categories selectable via `tools` URL parameter, `--tools` flag or `TOOLS` env var
- Hosted OAuth endpoint plus local stdio mode and an .mcpb bundle for supported clients
- Also covers Actor builds, schedules, tasks, storage and Apify API access as opt-in tools
Weaknesses
- Requires an Apify account; Actors run and store results on Apify's platform, not locally
- Actor runs are paid per Actor pricing; cost depends on each Actor
- call-actor waits only 30 seconds by default, up to 45; longer runs need polling
- Local stdio mode lacks inferred output schemas for Actor results
- no GPU
- Docker
- Needs Apify account and API token or OAuth, Apify platform (Actors), Node.js 22+ for local stdio
xiaohongshu-mcp
A Go MCP server that drives a headless browser to operate a logged-in Xiaohongshu account. It exposes 13 tools over streamable HTTP at /mcp: publish image or video posts (with tags, scheduling, visibility), search with filters, read feeds, post details and comments, reply, like, favorite and fetch user profiles. Works with Claude Code, Cursor, VS Code, Gemini CLI and Cline.
Who it is for: Developers connecting an AI assistant to Xiaohongshu for posting and research
Strengths
- Covers publishing, search, comments, likes, favorites and profiles as MCP tools
- Prebuilt binaries for macOS arm64, Windows x64 and Linux x64, plus Docker images
- Optional Bearer token auth via AUTH_TOKEN and optional proxy via XHS_PROXY
- README gives config snippets for Claude Code, Cursor, VS Code, Gemini CLI and Cline
Weaknesses
- Automates the web UI, so site changes or risk controls can break posting
- One account cannot stay logged in on other web clients without kicking the MCP session
- No prebuilt binary for macOS Intel or Linux ARM64
- Auth is off by default; the noVNC GUI image has no password and must not be public
- no GPU
- Docker + Compose
- Needs Headless browser (about 150MB, auto-downloaded on first run), Xiaohongshu account
- port 18060
More in MCP servers
Building with these?
We help teams pick an open-source AI stack and run it in production. Client work never changes a rank.