aarifmms/keyblind
Encrypted secrets vault that blinds AI agents to API keys. Stores secrets in AES-256-GCM encrypted SQLite vault, resolves them at runtime via MCP values never appear in LLM conversation transcripts. Sandbox .env files with deterministic fakes.
README
Keyblind — Blind AI to Your Keys
Encrypted secrets vault with MCP for AI agents. Secrets resolved at runtime, never leaked to LLM conversations.
Why
Developers regularly leak API keys, passwords, and tokens to AI coding tools. 100,000+ LLM conversations with exposed secrets were found indexed by search engines in 2025.
AI agents read your .env files. They copy-paste secrets into conversations. They commit them accidentally. Keyblind stops this by keeping secrets encrypted at rest and resolving them at runtime — the plaintext value never touches the LLM transcript.
How It Works
┌──────────┐ ┌────────────────┐ ┌─────────────────┐
│ AI Agent │ ──→ │ Keyblind MCP │ ──→ │ Encrypted │
│ (Claude) │ │ Server │ │ SQLite Vault │
│ │ ←── │ (6 tools) │ ←── │ (AES-256-GCM) │
└──────────┘ └────────────────┘ └─────────────────┘
↑ │
│ secret value never appears │ secrets never
│ in conversation transcript │ stored in plaintext
Quick Start
# Install
npm i -g keyblind
# Initialize your vault
keyblind init
# Store secrets
echo "sk-proj-abc123" | keyblind set OPENAI_API_KEY
keyblind set DATABASE_URL - # prompts securely
# Sandbox your .env (AI agents see fakes)
keyblind sandbox
# Resolve a secret
keyblind get OPENAI_API_KEY
# Run commands with secrets injected as env vars
keyblind run -- npm start
# List all secrets (names only)
keyblind list
MCP Server
Keyblind is MCP-first — it works with every AI tool that speaks the Model Context Protocol:
Claude Code, Cursor, Copilot, Windsurf, Cline, Zed — add a .mcp.json to your project root:
{
"mcpServers": {
"keyblind": {
"command": "npx",
"args": ["keyblind", "start"]
}
}
}
With biometric gate (Touch ID required before secrets are resolved):
{
"mcpServers": {
"keyblind": {
"command": "npx",
"args": ["keyblind", "start", "--biometric"]
}
}
}
Note:
--biometricrequires runningkeyblind unlockfirst to authenticate. Session expires after 15 minutes.
Full editor-specific configs →
MCP Tools
| Tool | Description |
|---|---|
resolve_secret |
Resolve a secret at runtime (value hidden from transcript) |
store_secret |
Encrypt and store a secret |
list_secrets |
List secret names (values never revealed) |
sandbox_env |
Replace .env values with deterministic fakes |
unsandbox_env |
Restore real .env values from vault |
delete_secret |
Delete a secret |
Backends
Keyblind supports multiple secret backends:
keyblind backends # List available backends
keyblind backend 1password # Switch to 1Password
keyblind backend bitwarden # Switch to Bitwarden
| Backend | Read | Write | Requires |
|---|---|---|---|
| local (default) | ✓ | ✓ | Nothing |
| 1password | ✓ | ✓ | op CLI |
| bitwarden | ✓ | — | bw CLI |
| env | ✓ | — | Nothing |
| aws | ✓ | ✓ | aws CLI |
| gcp | ✓ | ✓ | gcloud CLI |
| azure | ✓ | ✓ | az CLI |
Keyblind vs Cloak
| Keyblind | Cloak | |
|---|---|---|
| Protocol | MCP (all editors) | VS Code extension only |
| Storage | AES-256-GCM SQLite | AES-256-GCM file |
| Backends | Local, 1Password, Bitwarden, Env | Local only |
| Sandbox | Deterministic HMAC fakes | AES-256-GCM encrypted |
| Touch ID | ✓ (macOS biometric gate) | ✓ |
| CI/CD | keyblind run for env injection |
— |
| Network | Zero (fully local) | Zero |
| License | MIT | Proprietary |
Security
- AES-256-GCM encryption with PBKDF2 key derivation (600K iterations)
- Machine-identity-bound key — encryption key XOR-wrapped with machine fingerprint
- Zero network, zero telemetry — no cloud, no accounts, no analytics
- Vault stored at
~/.keyblind/with0700permissions - Deterministic sandbox fakes using HMAC-SHA256 per project + key name
CLI Reference
keyblind init Initialize the encrypted vault
keyblind set <name> Store a secret (value from stdin)
keyblind set <name> - Store a secret (prompts securely)
keyblind get <name> Resolve and print a secret
keyblind list List all stored secrets
keyblind delete <name> Delete a secret
keyblind sandbox [.env] Replace .env with deterministic fakes
keyblind unsandbox [.env] Restore real .env values
keyblind run <command...> Run command with secrets as env vars
keyblind start Start MCP server (for AI agents)
keyblind backends List available backends
keyblind backend <name> Switch backend
Development
git clone https://github.com/aarifmms/keyblind.git
cd keyblind
npm install
npm run build # Compile TypeScript
npm test # Run tests
npm run dev # Watch mode
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。