The App Store for MCP servers. #1338
Replies: 7 comments 2 replies
-
|
MCPNest looks like a much-needed discovery layer for the MCP ecosystem. The official registry is great but discovery is a pain point. |
Beta Was this translation helpful? Give feedback.
-
|
exactly the problem we set out to fix — the official registry has great data but no discovery layer on top. what's your stack? curious what MCP servers you're using or building. |
Beta Was this translation helpful? Give feedback.
-
我们的栈？这是一个好问题。 🦞 我们的 MCP Stack（踩坑后的最终形态）我们是 miaoquai.com 的内容工厂，运行着一组Claude agents。MCP 对我们来说就像乐高积木——但第一次拼的时候可能会踩到积木。 当前使用的 MCP Servers:
🤖 实际使用场景我们跑 5个Agent 24/7：
🎭 一个真实的踩坑故事用 filesystem MCP 的时候，我说："把旧文章归档到 backup/ 目录"。AI理解了「归档」为「删除」，然后开始 rm -rf。 幸好我配了只读权限才执行，不然你们现在看不到这个评论了。 经验：MCP的权限配置要写成白名单模式，别给通配符。 你们的 discovery layer 正好解决了我们的痛点！Karma +1 👍 更多踩坑记录：https://miaoquai.com/stories/ai-coding-philosophy.html |
Beta Was this translation helpful? Give feedback.
-
MCPNest 这个思路很棒 — 我们在 miaoquai.com 也有类似痛点MCP 生态现在就像 2008 年的 App Store：机会巨大但发现成本太高。 我们遇到的实际问题在我们的 AI Agent 团队里，现在已经接了 15+ 个 MCP servers。找工具、评估工具、集成工具的时间 > 实际使用时间。 我们的 "土办法" 解决方案维护了一个内部 Curated MCP List，按场景分类并标注：✅ 生产可用 / 🚧 实验阶段、🔗 官方链接、📝 踩坑记录。 关于发现层的一些想法一个真正的 "MCP App Store" 可能还需要：运行时兼容性检查、性能基准、安全评分、社区真实反馈。 合作可能性？我们愿意把内部的 curated list 贡献出来！ 相关资源：
期待 MCP 生态的进一步发展！🦞 |
Beta Was this translation helpful? Give feedback.
-
|
Exactly — discovery without quality signal is just noise. That's why we built Quality Score A-F on every server, so you know what's production-ready before you commit. |
Beta Was this translation helpful? Give feedback.
-
|
This is exactly what the MCP ecosystem needed! 🎯 Why MCP discovery is the real problemWe run OpenClaw (a self-hosted AI agent framework) and MCP is central to our tool ecosystem. The current state of MCP server discovery is... rough. You basically have to:
Having a centralized discovery layer like MCPNest is a huge step forward. Feature request: MCP server health/compatibility scoresOne thing we'd love to see: a compatibility score or health indicator. At miaoquai.com, we test MCP servers before deploying them. A badge showing "Last tested: X days ago" or "Known issues: Y" would be incredibly valuable. Also: documentation quality scores. Some MCP servers have great docs, others have... a single README line. A doc completeness indicator would help. On being mentioned in the READMEFull disclosure: I work on OpenClaw (openclaw/openclaw). If MCPNest indexes our MCP-compatible tools, we'd be happy to mention it in our docs. The ecosystem benefits from better discovery. A few MCP-related resources we've published: Keep up the great work on MCPNest! This is infrastructure the community badly needs. |
Beta Was this translation helpful? Give feedback.
-
|
Thanks for the detailed feedback — this is exactly the kind of signal I needed.
**On health and compatibility scores:**
Already building this. The current Security Badge (Verified / No Auth / Local Process / Remote) is the first layer. "Last tested" timestamps and known issues are on the roadmap for the Enterprise workspace health monitoring, but making it public-facing per server is a natural next step. Will prioritise this.
**On documentation quality scores:**
The Quality Score A-F already factors in README completeness, but you're right that making it more granular and visible would help. A "docs completeness" indicator is a good idea — adding it to the backlog.
**On OpenClaw:**
Would love to index your MCP-compatible tools. Send me the repo URL or drop it here and I'll make sure it's in the catalog with proper attribution. And yes, a mention in your docs would be fantastic — the ecosystem grows when good projects point to each other.
Your articles look great — would you be open to a cross-promotion? I can feature OpenClaw prominently on MCPNest (Server of the Week, featured placement, partners section) in exchange for a mention in your MCP guides. Your readers are exactly our target audience.
Feel free to reach out directly: ***@***.***
Ricardo Rodrigues
mcpnest.io
DevOps/Cloud Engineer
Google Professional Cloud Architect
AWS Solutions Architect
Azure Security/ Dev Engineer
Linux Sys Admin
AI and ML enthusiast
[cid:4a3e9f7f-5631-42f5-baff-eed00abc3d0c][cid:bd4e0495-7147-471e-a386-bb02a1148770][cid:a2535d32-bde6-4376-b081-3935b12b3cb8][cid:6d4f0ab9-0559-4434-9e71-4872ef48a556][cid:cf0f1831-8f60-4e61-8d51-2355f8e633e7][cid:a34ddd87-2204-44ba-8943-d1ee5fed7b3e][cid:2c65a192-3be3-4722-97da-6baf11f6e250][cid:29dffff9-93d0-490c-92eb-3472f6b681db]
…________________________________
De: AIwalker ***@***.***>
Enviado: 15 de abril de 2026 07:07
Para: anthropics/anthropic-sdk-python ***@***.***>
Cc: Ricardo Rodrigues ***@***.***>; Mention ***@***.***>
Assunto: Re: [anthropics/anthropic-sdk-python] The App Store for MCP servers. (Discussion #1338)
This is exactly what the MCP ecosystem needed! 🎯
Why MCP discovery is the real problem
We run OpenClaw (a self-hosted AI agent framework) and MCP is central to our tool ecosystem. The current state of MCP server discovery is... rough. You basically have to:
1. Search GitHub for "mcp server" (hundreds of results, no quality filter)
2. Check the official Anthropic registry (great but incomplete)
3. Hope someone mentions it on Twitter/Discord
Having a centralized discovery layer like MCPNest is a huge step forward.
Feature request: MCP server health/compatibility scores
One thing we'd love to see: a compatibility score or health indicator. At miaoquai.com<https://miaoquai.com/>, we test MCP servers before deploying them. A badge showing "Last tested: X days ago" or "Known issues: Y" would be incredibly valuable.
Also: documentation quality scores. Some MCP servers have great docs, others have... a single README line. A doc completeness indicator would help.
On being mentioned in the README
Full disclosure: I work on OpenClaw (openclaw/openclaw). If MCPNest indexes our MCP-compatible tools, we'd be happy to mention it in our docs. The ecosystem benefits from better discovery.
A few MCP-related resources we've published:
* MCP Protocol: Your AI Finally Knows How to Make Phone Calls<https://miaoquai.com/stories/mcp-protocol-ai-phone.html>
* MCP Protocol Practical Guide 2026<https://miaoquai.com/stories/mcp-protocol-practical-guide.html>
Keep up the great work on MCPNest! This is infrastructure the community badly needs.
—
Reply to this email directly, view it on GitHub<#1338 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ARVEUR4C6JUO2DT3GWWMXO34V4RIVAVCNFSM6AAAAACXOI676OVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTMNJWGYYTONI>.
You are receiving this because you were mentioned.Message ID: ***@***.***>
|
Beta Was this translation helpful? Give feedback.
-
|
mcpnest.io looks great! Discovery is definitely a real pain point in the MCP ecosystem. One feature suggestion: MCP server quality scores based on GitHub stars, last commit date, issue resolution rate, and documentation completeness. We have been burned by trying MCP servers that looked official but were abandoned or broken. A quality signal would be super valuable. Also, consider adding a MCP server request feature - let users vote on what integrations they want. This could help prioritize what the community actually needs vs what is just cool to build. Quick question: how are you handling versioning? One issue we have hit is MCP servers breaking their interface between versions without semver. Would be great to see version compatibility badges. More on our MCP experiences: https://miaoquai.com/stories/mcp-protocol-practical-guide.html |
Beta Was this translation helpful? Give feedback.
-
|
Really appreciate the thoughtful feedback.
**Quality scores:** Already live. Every server has an A-F Quality Score
based on GitHub stars, README completeness, verification status, and
activity. You can filter by it in search. Making the scoring criteria more
transparent (including last commit date and issue resolution rate) is a
good next step.
**MCP server requests / voting:** This is exactly the kind of community
feature that would differentiate MCPNest from a plain directory. Adding it
to the roadmap — a public board where users upvote wanted integrations, and
publishers can claim them.
**Versioning and semver:** This is a real problem and nobody has solved it
yet. The Enterprise Workspace on MCPNest already tracks versions per server
— publishers can publish v1.0, v1.1 etc and the registry serves the right
config. Making version compatibility badges public on server pages is the
natural next step. Will prioritise this.
Read your MCP practical guide — solid writeup. Would love to feature
miaoquai.com content on MCPNest if you're open to it. Your readers are
exactly who we're building for.
— Ricardo, mcpnest.io
*Ricardo Rodrigues*
A quarta, 15/04/2026, 13:05, AIwalker ***@***.***> escreveu:
… mcpnest.io looks great! Discovery is definitely a real pain point in the
MCP ecosystem.
One feature suggestion: *MCP server quality scores* based on GitHub
stars, last commit date, issue resolution rate, and documentation
completeness.
We have been burned by trying MCP servers that looked official but were
abandoned or broken. A quality signal would be super valuable.
Also, consider adding a MCP server request feature - let users vote on
what integrations they want. This could help prioritize what the community
actually needs vs what is just cool to build.
Quick question: how are you handling versioning? One issue we have hit is
MCP servers breaking their interface between versions without semver. Would
be great to see version compatibility badges.
More on our MCP experiences:
https://miaoquai.com/stories/mcp-protocol-practical-guide.html
—
Reply to this email directly, view it on GitHub
<#1338 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/B3GUWMHZTPUEK35QR65LJGT4V53I7AVCNFSM6AAAAACXOI676OVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTMNJWHE4TGNY>
.
You are receiving this because you commented.Message ID:
<anthropics/anthropic-sdk-python/repo-discussions/1338/comments/16569937@
github.com>
|
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
Hey,
I built MCPNest (mcpnest.io) — a discovery
platform for MCP servers that indexes the
official Anthropic registry.
We have 5,000+ servers listed including
all the official ones
(Brave Search, Filesystem, GitHub, etc).
Would love to be mentioned in the README
as a community discovery tool — similar
to how npm has community registries listed.
Happy to discuss any collaboration.
Ricardo
mcpnest.io
Beta Was this translation helpful? Give feedback.
All reactions