Kratos MCP
Ultra-lean memory system for AI coding tools that stores project knowledge locally with SQLite and enables AI to remember your project across sessions.
README
<div align="center">
Kratos MCP
Ultra-Lean Memory System for AI Coding Tools
Never explain your codebase again. Let AI remember everything.
</div>
<div align="center">
Kratos has evolved. This repo is now legacy.
github.com/ceorkm/kratos-cli
Kratos is now a standalone CLI — faster, lighter, works with ANY AI agent.
No MCP overhead. No vendor lock-in. Just memory that works.
npx kratos-memory
Claude Code, Codex, Cursor, Cline — anything that runs Bash.
Your existing memories work instantly. Same ~/.kratos/ storage, same SQLite format.
Go to Kratos CLI >
</div>
Why Kratos?
After building 30+ production apps with AI, we discovered a critical problem: AI tools forget everything between sessions. You explain your architecture, your patterns, your decisions—and tomorrow, you explain it all again.
Kratos MCP solves this with an ultra-lean memory system that gives AI perfect recall of your project—with minimal context overhead.
Features
<table> <tr> <td width="50%">
100% Project Isolation
Each project gets its own SQLite database. No cross-contamination. Ever.
</td> <td width="50%">
Zero Configuration
Auto-detects projects via git, package.json, or directory structure. Just install and code.
</td> </tr> <tr> <td width="50%">
Ultra-Lean Architecture
Just 12 essential tools. 64% smaller context footprint than competitors.
</td> <td width="50%">
Universal Protocol
Works with Claude, Cursor, Windsurf, Continue—any MCP-compatible tool.
</td> </tr> </table>
Installation
# Install globally
npm install -g kratos-mcp
# Or run directly with npx (no installation required)
npx kratos-mcp
# Or install as a dependency
npm install kratos-mcp
Quick Start
1. Configure Your AI Tool
<details> <summary><b>Claude Desktop</b></summary>
Add to your Claude Desktop config file:
- macOS:
~/Library/Application Support/Claude/claude_desktop_config.json - Windows:
%APPDATA%\Claude\claude_desktop_config.json - Linux:
~/.config/claude/claude_desktop_config.json
{
"mcpServers": {
"kratos": {
"command": "npx",
"args": ["--yes", "kratos-mcp@latest"]
}
}
}
Or if you have it installed globally:
{
"mcpServers": {
"kratos": {
"command": "kratos-mcp",
"args": []
}
}
}
</details>
<details> <summary><b>Claude Code</b></summary>
Run this command in your terminal:
claude mcp add kratos -- npx --yes kratos-mcp@latest
Or for global installation:
# First install globally
npm install -g kratos-mcp@latest
# Then add to Claude Code
claude mcp add kratos -- kratos-mcp
</details>
<details> <summary><b>Cursor</b></summary>
Add to .cursor/mcp_config.json in your project root:
{
"mcpServers": {
"kratos": {
"command": "npx",
"args": ["--yes", "kratos-mcp@latest"]
}
}
}
</details>
<details> <summary><b>Other MCP Tools</b></summary>
Kratos works with any tool supporting the Model Context Protocol. The general format is:
{
"command": "npx",
"args": ["kratos-mcp"]
}
Compatible with: Windsurf, Cline, BoltAI, Augment Code, Roo Code, Zencoder, Amazon Q, Qodo Gen, JetBrains AI, Warp, Opencode, Continue.dev, Zed, and more!
Check your tool's documentation for specific MCP server configuration location.
</details>
2. Start Using Kratos
// Your AI now remembers:
// - Your authentication patterns
// - Your API structure
// - Your component architecture
// - Your coding standards
// - Every decision you've made
Available Tools
Kratos provides 12 ultra-lean tools optimized for minimal context consumption:
Memory Management (7 tools)
| Tool | Description |
|---|---|
memory_save |
Store important project knowledge with tags, paths, and importance levels |
memory_search |
Smart semantic search with debug mode and path matching |
memory_ask |
Natural language queries about your memories |
memory_get_recent |
Get recently created memories with filtering |
memory_get |
Retrieve a specific memory by ID |
memory_get_multiple |
Bulk retrieve multiple memories |
memory_forget |
Delete a memory by ID |
Security (1 tool)
| Tool | Description |
|---|---|
security_scan |
Scan text for PII and secrets before saving |
Project Management (3 tools)
| Tool | Description |
|---|---|
project_switch |
Switch between different projects |
project_current |
Get current active project info |
change_storage_path |
Dynamically change storage location with automatic data migration |
System (1 tool)
| Tool | Description |
|---|---|
system_status |
Get system status and memory statistics |
How It Works
AI Tool ──► Kratos MCP ──► Project Detection ──► SQLite + FTS5 ──► Perfect Context
- SQLite + FTS5: Lightning-fast full-text search
- Smart Scoring: Path matching + recency + importance
- Auto-detection: Git, package.json, or directory-based
- Secure: All data stays local, no external calls
- Lean: Only 4 core components, minimal memory footprint
Performance
| Metric | Value |
|---|---|
| Context Overhead | 64% smaller than v3 |
| Memory Retrieval | < 10ms |
| Project Switch | < 100ms |
| Storage Overhead | ~2MB per project |
Data Storage
~/.kratos/
├── projects/
│ ├── project-id-1/
│ │ └── memories.db # SQLite database with FTS5
│ └── project-id-2/
│ └── memories.db
└── global/
└── global.db # Shared knowledge (optional)
Use change_storage_path to move data to custom locations like /opt/kratos or .kratos for per-project storage.
Example Usage
// Save a memory
await memory_save({
summary: "JWT auth implementation",
text: "We use httpOnly cookies with refresh tokens...",
tags: ["auth", "security"],
paths: ["src/middleware/auth.ts"],
importance: 5
});
// Search memories
await memory_search({
q: "authentication",
k: 5,
debug: true
});
// Ask natural language questions
await memory_ask({
question: "How does our auth system work?",
limit: 10
});
Contributing
git clone https://github.com/ceorkm/kratos-mcp.git
npm install
npm run build
npm run dev
License
MIT
<div align="center">
Built for developers who value their time.
</div>
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。
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。
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。
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。
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。
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。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
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。