SnapBack MCP Server
Enables AI-powered code safety analysis including risk detection, secret scanning, dependency checking, and code snapshot management. Works offline for basic features with optional cloud integration for advanced ML analysis and team collaboration.
README
SnapBack MCP Server
AI-powered code analysis and snapshot management via Model Context Protocol
Integrate SnapBack's code safety features directly into Claude Desktop, Cursor, and any MCP-compatible AI tool.
Quick Start
npm install -g @snapback/mcp-server
snapback-mcp
Works immediately - no configuration required!
Features
🆓 Free (No Account Needed)
- ✅ Risk Analysis: Detect secrets, vulnerabilities in code changes
- ✅ Dependency Checking: Validate package.json changes
- ✅ Local Analysis: Basic secret detection and security scanning
- ✅ Offline Mode: Works without internet connection
- ✅ Context7 Integration: Library documentation and code search
☁️ Pro Features (Optional API Key)
Get a free API key from snapback.dev to unlock:
- 🔐 Advanced ML Analysis: AI-powered risk detection
- 🔐 Snapshot Management: Create and restore code snapshots
- 🔐 Cloud Sync: Access snapshots across devices
- 🔐 Team Sharing: Collaborate on code safety
Installation
Claude Desktop
Add to ~/Library/Application Support/Claude/claude_desktop_config.json:
{
"mcpServers": {
"snapback": {
"command": "npx",
"args": ["-y", "@snapback/mcp-server"]
}
}
}
With API Key (Optional)
{
"mcpServers": {
"snapback": {
"command": "npx",
"args": ["-y", "@snapback/mcp-server"],
"env": {
"SNAPBACK_API_KEY": "your_api_key_here"
}
}
}
}
Cursor / Other MCP Clients
# Install globally
npm install -g @snapback/mcp-server
# Run with stdio transport
snapback-mcp
Available Tools
snapback.analyze_risk
Analyze code changes for potential security risks before applying them.
When to use:
- Before accepting AI-generated code
- When reviewing complex changes
- For critical files (auth, database, config)
Example:
// AI detects you want to add authentication
// Before applying changes, it calls:
snapback.analyze_risk({
changes: [
{ added: true, value: "const API_KEY = 'sk_live_...';" }
]
})
// Returns: ⚠️ HIGH RISK: Hardcoded secret detected
snapback.check_dependencies
Check for dependency-related risks when package.json changes.
Example:
snapback.check_dependencies({
before: { "lodash": "^4.17.15" },
after: { "lodash": "^4.17.21" }
})
// Returns: ℹ️ Security update available
snapback.create_snapshot (Pro)
Create a code snapshot before risky changes.
Example:
snapback.create_snapshot({
reason: "Before major refactor",
files: ["src/auth.ts", "src/db.ts"]
})
// Returns: ✅ Snapshot created: snap_xyz123
snapback.list_snapshots (Pro)
List all available snapshots.
snapback.restore_snapshot (Pro)
Restore code from a previous snapshot.
Context7 Tools
ctx7.resolve-library-id: Find library documentationctx7.get-library-docs: Fetch library docs and examples
Configuration
Environment Variables
# Optional: SnapBack API key for Pro features
SNAPBACK_API_KEY=sk_...
# Optional: Custom API URL
SNAPBACK_API_URL=https://api.snapback.dev
# Optional: Context7 API key for enhanced docs
CONTEXT7_API_KEY=...
# Optional: Log level
LOG_LEVEL=info
Offline Mode
Works perfectly without any configuration or API keys:
# No env vars needed!
npx @snapback/mcp-server
What works offline:
- Risk analysis (basic)
- Dependency checking
- Secret detection
- Context7 library search (cached)
What requires API key:
- Advanced ML risk analysis
- Snapshot creation/restoration
- Cloud sync
- Team features
Architecture
┌─────────────────┐
│ AI Tool │ (Claude, Cursor, etc.)
│ (MCP Client) │
└────────┬────────┘
│ MCP Protocol
│
┌────────▼─────────────────────────────────┐
│ SnapBack MCP Server │
│ ┌──────────────┐ ┌──────────────────┐ │
│ │ Free Tools │ │ Pro Tools │ │
│ │ - analyze │ │ - snapshots │ │
│ │ - check_deps │ │ - cloud sync │ │
│ └──────────────┘ └──────────────────┘ │
└───────────┬──────────────────────────────┘
│
┌───┴────┐
│ │
┌────▼───┐ ┌─▼─────────┐
│ Local │ │ SnapBack │
│Analysis│ │ API │
└────────┘ └───────────┘
Development
Running Locally
git clone https://github.com/snapback-dev/mcp-server.git
cd mcp-server
pnpm install
pnpm build
pnpm start
Testing
# Run tests
pnpm test
# Test without API key (offline mode)
unset SNAPBACK_API_KEY
pnpm start
# Test with API key
export SNAPBACK_API_KEY=sk_test_...
pnpm start
Building
pnpm build
# Output: dist/index.js (ESM)
Troubleshooting
Server won't start
- Check Node.js version:
node -v(requires 18+) - Clear cache:
rm -rf node_modules && npm install - Check permissions:
chmod +x $(which snapback-mcp)
API key not working
- Verify key format:
sk_live_...orsk_test_... - Check env var:
echo $SNAPBACK_API_KEY - Get new key: snapback.dev/settings/api
Tools not showing in Claude
- Restart Claude Desktop completely
- Check config file syntax (JSON must be valid)
- Look for errors in Claude's console logs
Security
- All secrets handled via environment variables
- No data sent to SnapBack without API key
- Local analysis runs offline
- Open source - audit the code yourself
Report security issues: security@snapback.dev
Links
- Documentation: docs.snapback.dev
- Main Repository: Marcelle-Labs/snapback.dev
- Issues: github.com/snapback-dev/mcp-server/issues
- NPM: @snapback/mcp-server
License
Apache-2.0 © SnapBack
Related
snapbackVS Code Extension@snapback/sdk- TypeScript SDK@snapback/contracts- Type definitions
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。