AI-Collab-Memory
An MCP server for managing work logs, research results, and task checkpoints to enable seamless collaboration and state recovery between AI agents. It provides a persistent memory layer for tracking project history and resuming workflows across different sessions or tools.
README
🧠 AI-Collab-Memory
AI Collaboration Memory - MCP server for work logs, research sharing, and checkpoint recovery
⚠️ What This Is (and Isn't)
| ✅ Can Do | ❌ Cannot Do |
|---|---|
| Save work logs | Real-time AI communication |
| Share research results | Auto command delivery |
| Checkpoint (save point) for recovery | Wake up other AIs to execute |
| Track who did what |
For real-time collaboration? → Use Antigravity to call Claude Code/Codex directly via terminal!
🚀 Installation
Prerequisites
- Node.js (v18+) - Download
Install Steps
- Clone or download this repository
- Double-click
install.bat(Windows) or run:npm install && npm run build - Done!
🔌 MCP Registration
💡 Replace
<install-path>with your actual installation path.
Claude Code
claude mcp add ai-collab-memory -- node "<install-path>/AI-Collab-Memory/dist/index.js"
Codex (~/.codex/config.toml)
[mcp_servers.ai-collab-memory]
command = "node"
args = ["<install-path>/AI-Collab-Memory/dist/index.js"]
Antigravity / Claude Desktop (JSON)
{
"mcpServers": {
"ai-collab-memory": {
"command": "node",
"args": ["<install-path>/AI-Collab-Memory/dist/index.js"]
}
}
}
🛠️ Tools (7)
Work Log
| Tool | Description |
|---|---|
log_work |
Save work record |
get_work_log |
View work logs |
Research
| Tool | Description |
|---|---|
save_research |
Save research results |
search_research |
Search research |
Checkpoint (Save Point)
| Tool | Description |
|---|---|
save_checkpoint |
Save current state (for recovery) |
load_checkpoint |
Load saved state |
Status
| Tool | Description |
|---|---|
memory_status |
View overall status |
📖 Usage Examples
Log work
"Log this work. Claude Code created the API."
Save research
"Save this research result."
Save checkpoint (before interruption)
"Save checkpoint now. I'll continue tomorrow."
Resume work
"Load the last checkpoint."
Check status
"Show memory status."
💡 Key Use Cases
Case 1: Recovery After Interruption
1. During work
→ "Save checkpoint"
2. (Computer shuts down / Session disconnects)
3. Next day
→ "Load last checkpoint"
→ See progress + next steps
→ Resume immediately!
Case 2: Track Collaboration History
1. Claude Code: Finished API development
→ "Log this work"
2. Codex: Frontend development
→ "Log this work"
3. Later
→ "Show work log"
→ See who did what at a glance
Case 3: Share Research
1. Antigravity: Research React 18
→ "Save research"
2. Claude Code: When writing code
→ "Search React research"
→ Use already researched content!
📁 Project Structure
AI-Collab-Memory/
├── src/
│ └── index.ts ← MCP server source (7 tools)
├── data/ ← Runtime data (auto-created, git-ignored)
│ ├── work_log.json
│ ├── research.json
│ └── checkpoint.json
├── dist/ ← Build output (auto-generated)
├── install.bat ← Windows one-click installer
├── package.json
├── tsconfig.json
├── SKILL.md ← Claude Skill definition
├── README.md ← English docs (this file)
├── README.ko.md ← Korean docs
└── LICENSE
📄 License
MIT License. See LICENSE for details.
👤 Author
Sodam AI Studio
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。
mcp-server-qdrant
这个仓库展示了如何为向量搜索引擎 Qdrant 创建一个 MCP (Managed Control Plane) 服务器的示例。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。