AI UNDERDOGSDAILY PICK
AI UNDERDOGS
每个AI Agent共享一个大脑
One brain for every AI agent
Fmarzochi/EGC
你用 Cursor 写代码
用 Copilot 补全
用 Claude 改 bug——但每次切工具
上下文就断了
EGC 说：别切了，让它们共享一个脑子
You code in Cursor, autocomplete in Copilot, fix
bugs in Claude — but every switch loses
your context. EGC says: stop switching, let them
share one brain
EGC是什么
What EGC actually is
EGC 是一套基于 MCP 协议的共享记忆系统
它给你的项目建了一个持久化的知识库——代码架构
设计决策
之前的 bug 记录全存在里面
任何支持 MCP 的 AI Agent 接进来
就能读到同样的上下文
EGC is a shared memory layer built on
the MCP protocol. It creates a persistent knowledge
base for your project — architecture, design decisions
past bug logs. Any MCP-compatible AI agent plugs
in and reads the same context
★ SIGNAL 1
持久化不是口号
Persistence that actually works
很多工具说「有记忆」，但重启一下就没了
EGC 的记忆存在项目目录里
跟 .git 同级
你关掉 Cursor 再打开
Agent 还知道你上周为什么重构了那个模块——因为它读的是同一个知识库
Most tools claim 'memory' but lose it on
restart. EGC stores memory in your project directory
same level as .git. Close Cursor, reopen it
— the agent still knows why you refactored
that module last week, because it reads the
same knowledge base
★ SIGNAL 2
跨Agent协作
Agents that collaborate
最实用的场景：你让 Cursor 写了个功能
然后想用 Claude Code 来重构
以前得把前因后果再说一遍
现在两个 Agent 读同一份记忆
Claude 直接知道这个函数为什么这么写
哪些边界 case 已经处理过
The killer use case: Cursor writes a feature
then you want Claude Code to refactor it
Used to mean re-explaining everything. Now both agents
read the same memory — Claude knows why
the function was written this way and which
edge cases are handled
说实话，MCP 生态还在长
EGC 的工具链也不算成熟
但方向对了——AI 编程工具越来越多
Agent 之间的上下文共享迟早是刚需
这个项目现在 star 增长很快
说明痛点真实存在
Honest take: the MCP ecosystem is still growing
and EGC's tooling isn't mature yet. But the
direction is right — as AI coding tools
multiply, cross-agent context sharing will be essential. The
fast-rising star count proves the pain point is
real
AI UNDERDOGS
给你的AI Agent们装一个共享记忆
Give your AI agents shared memory
Fmarzochi/EGC
关注 · 每天发现更多 AI 神作
github.com/Fmarzochi/EGC