lockbox-mcp
Enables AI coding agents to securely store and retrieve encrypted API keys via MCP tools.
README
Lockbox
Encrypted API key vault with MCP integration for AI coding tools.
Lockbox stores your API keys in an AES-256-GCM encrypted vault on your machine. Access them from the CLI, pipe them into scripts, or let AI coding agents (Claude Code, Cursor, Windsurf) retrieve secrets through the Model Context Protocol — without ever putting real keys in your codebase.
Quick Start
npm install -g lockbox-vault
lockbox init # create your encrypted vault
lockbox add openai API_KEY sk-... # store a secret
lockbox get openai API_KEY # retrieve it
Installation
npm install -g lockbox-vault
Requires Node.js 18+.
CLI Commands
Vault Lifecycle
lockbox init # create a new encrypted vault
lockbox unlock # unlock with your master password (15-min session)
lockbox lock # lock the vault immediately
lockbox status # show vault info, lock state, key count
lockbox doctor # run health checks on the vault
Secret Management
lockbox add openai API_KEY sk-abc123 # store a secret
lockbox add stripe SECRET --project myapp # organize by project
lockbox add aws ACCESS_KEY # prompts for value (hidden)
lockbox get openai API_KEY # print to stdout
lockbox get openai API_KEY --copy # copy to clipboard (clears in 30s)
lockbox remove openai API_KEY # delete (with confirmation)
lockbox list # show all keys (never values)
lockbox list --project myapp # filter by project
lockbox search stripe # search by name, service, or notes
Import & Export
lockbox import .env # import from .env file
lockbox import creds.json --project myapp # import from JSON
lockbox export > .env # export as .env format
lockbox export --format json # export as JSON
lockbox export --format shell >> ~/.bashrc # export as shell exports
Proxy Key System
The proxy key system lets you commit .env files to git without exposing real secrets. Instead of actual values, your env file contains lockbox:// references that are resolved at runtime.
# generate a .env.proxy file (safe to commit)
lockbox proxy-init --project myapp
# the generated file looks like:
# OPENAI_API_KEY=lockbox://openai/API_KEY
# STRIPE_SECRET_KEY=lockbox://stripe/SECRET_KEY
# run your app with real values injected
lockbox run --env-file .env.proxy "npm start"
lockbox run --env-file .env.proxy "python app.py"
Non-proxy values (like DATABASE_URL=postgresql://localhost/mydb) are passed through unchanged.
Audit & Security
lockbox audit # show last 50 audit entries
lockbox audit --since 2025-01-01 # filter by date
lockbox audit -n 10 # limit entries
MCP Setup
Lockbox exposes your vault to AI coding agents via the Model Context Protocol. The MCP server provides 6 tools: store_secret, get_secret, list_secrets, delete_secret, export_env, and list_projects.
Prerequisites: Unlock your vault before using MCP tools:
lockbox unlock
Claude Code
Add to your project's .mcp.json:
{
"mcpServers": {
"lockbox": {
"command": "npx",
"args": ["lockbox-mcp"]
}
}
}
Cursor
Add to .cursor/mcp.json in your project root (or ~/.cursor/mcp.json for global):
{
"mcpServers": {
"lockbox": {
"command": "npx",
"args": ["lockbox-mcp"]
}
}
}
Windsurf
Add to ~/.codeium/windsurf/mcp_config.json:
{
"mcpServers": {
"lockbox": {
"command": "npx",
"args": ["lockbox-mcp"]
}
}
}
Security Model
| Layer | Implementation |
|---|---|
| Encryption | AES-256-GCM with 12-byte random IV (fresh on every save) |
| Key Derivation | Argon2id (64 MB memory, 3 iterations, 4 threads) |
| Integrity | HMAC-SHA256 verified before decryption attempt |
| Session | Derived key encrypted to temp file, auto-expires after 15 minutes |
| Rate Limiting | 5 failed unlock attempts per 60s triggers 60s lockout |
| Audit Trail | Append-only log of all operations (values never logged) |
| Clipboard | Auto-clears after 30 seconds when using --copy |
| Memory | Derived key zeroed on lock; secrets only in memory while unlocked |
Vault File Format
The vault is a single encrypted JSON file at ~/.config/lockbox/vault.enc:
{ salt, iv, tag, ciphertext, hmac } (all base64-encoded)
The decrypted payload contains versioned key entries with per-key metadata (project, notes, timestamps, expiry).
Contributing
Contributions are welcome! Please open an issue or submit a pull request on GitHub.
git clone https://github.com/lockboxdev/lockbox.git
cd lockbox
npm install
npm run build
npm test
License
MIT
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。