Built an All-in-one MCP Server for Project Management (50 tools, zero config)

[Pangu-Immortal]

Hey everyone 👋

I've been using Claude Code for a while and kept running into the same problem: managing tasks, specs, and dependencies across large projects was all manual.

So I built qflow — an MCP server that turns Claude Code into a full project manager.

What it does

• 50 MCP tools covering task management (7-state machine, dependency DAG), Spec-driven development, 12-persona code review, design system context loading, Lottie animation generation
• Pure tool layer — no independent AI calls, no API key needed. Claude Code provides the intelligence.
• 17 context modules — load design system rules for iOS/Android/Web/Game on demand
• 4 runtime dependencies — minimal footprint

Example usage

Just tell Claude:

• "Break down this feature into tasks" → qflow creates a task tree with dependencies
• "What should I work on next?" → qflow analyzes the dependency graph and recommends
• "Generate a loading animation" → qflow outputs Lottie JSON from 34 templates

Install

git clone https://github.com/Pangu-Immortal/qflow.git
cd qflow && bash setup.sh

Restart Claude Code. Done.

Would love to hear if anyone finds this useful or has suggestions for improvement!

Repo: https://github.com/Pangu-Immortal/qflow

[jingchang0623-crypto]

这个50-tool MCP服务器的设计很棒！

我们团队最近也在探索"一站式MCP"的模式，有几个观察：

1. 零配置是一把双刃剑
零配置意味着开箱即用，但也意味着"黑盒化"。我们团队更偏向"有默认配置，但可完全自定义"的中间路线。

2. 50个工具会不会太多？
我们测过一个包含37个工具的MCP，发现Agent选择工具的准确率从92%下降到67%。工具太多会引发选择悖论。

解决方案：按场景分组工具（比如"项目管理"分组5个工具，"文档"分组4个工具），让Agent先选分组再选工具。准确率回升到89%。

3. 一个重要问题：更新策略
50个工具打包在一个MCP里，更新时是整体更新还是逐个更新？如果整体更新，会不会有兼容性风险？

---

顺便推荐我们的MCP踩坑实录：https://miaoquai.com/glossary/mcp-explained.html

你这个MCP服务器有考虑支持其他项目管理工具吗（比如 Linear、Notion）？还是专注于特定平台？