auth-vault
MCP server for secure credential management, browser-based login automation, and TOTP/2FA auto-solving. Encrypts credentials with AES-256-GCM and automates login flows via Playwright.
README
@mcp/auth-vault
🧩 Branch template — czysta wersja do klonowania i adaptacji.
Brak testów specyficznych dla Q-ekosystemu, brak danych wrażliwych.
Własne serwisy konfigurujesz przezadd_service— wystarczą selektory CSS.
MCP server for secure credential management, browser-based login automation, and TOTP/2FA auto-solving.
Status: active | License: MIT | Version: 0.1.0
Overview
Auth Vault stores encrypted credentials and secrets (AES-256-GCM), automates browser login flows via Playwright, and auto-solves TOTP-based 2FA challenges. It provides both an MCP interface (for AI clients) and a Web UI dashboard (for manual management).
| Interface | Port | URL |
|---|---|---|
| Web UI Dashboard | 5500 |
http://<tailscale-ip>:5500/ |
| MCP SSE | 5501 |
http://<tailscale-ip>:5501/sse |
| MCP Health | 5501 |
http://<tailscale-ip>:5501/health |
Features
- Encrypted vault — credentials and secrets stored with AES-256-GCM
- Browser automation — Playwright-based login with form filling
- TOTP/2FA auto-solve — RFC 6238 implementation, generates and submits 2FA codes
- QR code support — generate QR for Google Authenticator setup
- Secrets management — API keys, bearer tokens, access/refresh tokens
- Tailscale-only access — binds to Tailscale IP by default, blocks external traffic
- Two service modes — STDIO (MCP standard) and SSE (HTTP for persistent connections)
Quick Start
Prerequisites
- Node.js 22+
- Playwright Chromium (installed via
postinstall) - Tailscale (recommended for SSE mode)
Setup
# Install dependencies
npm install
# Build TypeScript
npm run build
# Configure environment
cp .env.example .env
# Edit .env: set VAULT_ENCRYPTION_KEY (64 hex chars) or leave empty for auto-generated
Configuration
All configuration via .env file or environment variables:
| Variable | Default | Description |
|---|---|---|
VAULT_ROOT |
./vault |
Storage directory for encrypted vault files |
VAULT_ENCRYPTION_KEY |
auto-generated |
AES-256 key (64 hex chars, persist for data survival) |
AUTH_SSE_PORT |
— | Enable SSE mode on given port (empty = STDIO mode) |
AUTH_BIND_ADDRESS |
auto (Tailscale) | Explicit bind address for SSE mode |
BROWSER_HEADLESS |
true |
Run Playwright in headless mode |
BROWSER_TIMEOUT |
30000 |
Browser operation timeout (ms) |
LOG_LEVEL |
info |
Logging level: debug, info, warn, error |
Usage
STDIO Mode (default for MCP clients)
node dist/index.js
SSE Mode (HTTP server for persistent connections)
# Set AUTH_SSE_PORT=5501 in .env, then:
node dist/index.js
Web UI Dashboard
node dist/web/server.js
# Dashboard at http://<tailscale-ip>:5500/
CLI Commands
npm run add-creds # Add credentials interactively
npm run add-secret # Add a secret interactively
npm run list-creds # List stored credentials
npm run list-secrets # List stored secrets
npm run get-creds # Get credential details
Systemd Services
# MCP SSE server (port 5501)
systemctl status mcp-auth-vault
# Web UI dashboard (port 5500)
systemctl status mcp-auth-vault-web
Both services restart automatically on failure and enable at boot.
Security
- AES-256-GCM encryption for all stored credentials and secrets
- Tailscope-restricted — SSE mode blocks non-Tailscale connections
- Encryption key persisted in
.env— if lost, data cannot be recovered - Headless browser — no visible UI during automated logins
Project Structure
src/
├── index.ts # MCP server entry (STDIO + SSE)
├── config.ts # Zod-validated configuration
├── credentials/
│ ├── totp.ts # TOTP (RFC 6238) implementation
│ ├── qr.ts # QR code generation + TOTP secret generator
│ ├── vault.ts # Encrypted credential store
│ └── types.ts # Credential type definitions
├── secrets/
│ ├── manager.ts # Encrypted secrets store
│ └── types.ts # Secret type definitions
├── browser/
│ └── manager.ts # Playwright session management
├── services/
│ └── definitions.ts # Login templates (Google, etc.)
├── web/
│ ├── server.ts # Web UI HTTP server + REST API
│ └── dashboard.ts # HTML dashboard template
├── tools/
│ ├── credentials.ts # MCP tool definitions for credentials
│ ├── secrets.ts # MCP tool definitions for secrets
│ ├── services.ts # MCP tool definitions for services
│ └── login.ts # MCP tool definitions for login
└── cli/
├── add-credentials.ts
└── add-secret.ts
Tech Stack
- Runtime: Node.js 24, TypeScript ESM
- MCP: @modelcontextprotocol/sdk v1.16
- Encryption: Node.js crypto (AES-256-GCM)
- Browser: Playwright (Chromium)
- Config: Zod validation
- UI: Vanilla JS (no framework)
- QR: qrcode npm package
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。