MCP 服务器怎么选？用 tashan score 看维护活跃度和真实采用量
Quick answer. 选 MCP 服务器别只看 star。tashan 用公开证据打 tashan score 分（维护活跃度 + 更新新鲜度，按真实下载量加权），并对每个能力做 deep / solid / thin 专家评级。下面是按 tashan score 排名的榜单，点开任意一行看完整安装方法和仓库健康度。
怎么判断一个 MCP 服务器还在维护、有没有人用？
看三件事：是否仍在维护（最近提交/发版）、有多少真实采用（npm 周下载）、以及是否有单一维护者/已归档风险。tashan 把这些合成一个可复现的 tashan score 分。
| # | Capability | tashan score | Verdict | Adoption |
|---|---|---|---|---|
| 1 | @supabase/mcp-server-supabase | 98.0 | solid | 112k/wk |
| 2 | @upstash/context7 | 98.0 | deep | 1.1M/wk |
| 3 | memory | 96.0 | — | 152k/wk |
| 4 | sequential-thinking | 94.0 | — | 109k/wk |
| 5 | chrome-devtools | 92.0 | deep | 3.3M/wk |
| 6 | everything | 92.0 | — | 243k/wk |
| 7 | filesystem | 90.0 | thin | 627k/wk |
| 8 | hostinger-api | 90.0 | deep | 119k/wk |
| 9 | @transcend-io/mcp-server-discovery | 87.0 | — | 149k/wk |
| 10 | @transcend-io/mcp-server-admin | 87.0 | — | 124k/wk |
| 11 | @transcend-io/mcp-server-preferences | 87.0 | — | 123k/wk |
| 12 | firecrawl | 86.0 | solid | 27k/wk |
MCP 服务器和 Agent Skill 有什么区别？
MCP 服务器通过 Model Context Protocol 给 AI 暴露工具；Agent Skill 是一个带 SKILL.md 的文件夹，按需加载。tashan 用同一套模型同时追踪并打分。
FAQ
这个榜单收钱排名吗？
不收。排名由公开证据算出——下载量、发版节奏和仓库健康度，没有付费位。
怎么安装？
点开任意能力页，复制对应客户端（Claude Code / Cursor / Claude Desktop / Codex）的安装片段即可。
tashan Pro$6/mo
This article and every score it links to are free, no account. Check your own config free too. Pro adds the series behind each row, and names a replacement when one is dying.
- The whole series behind any row, back to the first day we measured it
- The replacement, named — not just the news that something died
tashan doctorover your own config, on your own machine
Start a 7-day trial › Everything measured on this page stays free.