envibe

The missing permission layer between AI agents and your .env

The Problem

AI coding assistants (Claude Code, Cursor, Copilot) need your environment variables to run and test code. But they can see everything—API keys, database passwords, payment secrets.

It's all-or-nothing. Until now.

The Solution

envibe gives you per-variable access control with 5 levels:

| Level | AI Can See | AI Can Modify | Example | |-------|-----------|---------------|---------| | full | Actual value | Yes | NODE_ENV, PORT, DEBUG | | read-only | Actual value | No | DATABASE_URL | | placeholder | <VAR_NAME> | No | API_KEY | | schema-only | Format only | No | Complex configs | | hidden | Nothing | No | STRIPE_SECRET_KEY |

Quick Start

Add envibe as an MCP server to your AI tool:

{
  "mcpServers": {
    "envibe": {
      "command": "npx",
      "args": ["envibe-mcp"]
    }
  }
}

On first use, envibe automatically:

1. Creates .env.manifest.yaml from your .env.example
2. Generates .env.ai (filtered view for AI)
3. Blocks direct .env file access

How It Works

┌─────────────────────────────────────────────────────────────┐
│  Your .env (secrets)                                        │
│  ├── STRIPE_SECRET_KEY=sk_live_xxx    ← hidden from AI      │
│  ├── DATABASE_URL=postgres://...       ← AI can read        │
│  └── DEBUG=true                        ← AI can read/write  │
└─────────────────────────────────────────────────────────────┘
                           │
                           ▼
┌─────────────────────────────────────────────────────────────┐
│  .env.manifest.yaml (access rules)                          │
│  variables:                                                 │
│    STRIPE_SECRET_KEY: { access: hidden }                    │
│    DATABASE_URL: { access: read-only }                      │
│    DEBUG: { access: full }                                  │
└─────────────────────────────────────────────────────────────┘
                           │
                           ▼
┌─────────────────────────────────────────────────────────────┐
│  .env.ai (what AI sees)                                     │
│  DEBUG=true                    # [full]                     │
│  DATABASE_URL=postgres://...   # [read-only]                │
│  # STRIPE_SECRET_KEY hidden                                 │
└─────────────────────────────────────────────────────────────┘

Example Manifest

# .env.manifest.yaml
version: 1
variables:
  NODE_ENV:
    access: full
    description: "Environment mode"

  DATABASE_URL:
    access: read-only
    description: "Database connection string"

  OPENAI_API_KEY:
    access: placeholder
    description: "OpenAI API key"

  STRIPE_SECRET_KEY:
    access: hidden
    description: "Payment processing - never expose"

CLI Commands

| Command | Description | |---------|-------------| | envibe setup | Full setup (recommended) | | envibe setup -i | Interactive mode - choose access levels | | envibe generate | Regenerate .env.ai | | envibe view | Display variables with access levels | | envibe mcp | Start MCP server |

Installation

claude mcp add envibe npx envibe-mcp

Or add to .claude/settings.json:

{
  "mcpServers": {
    "envibe": {
      "command": "npx",
      "args": ["envibe-mcp"]
    }
  }
}

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "envibe": {
      "command": "npx",
      "args": ["envibe-mcp"]
    }
  }
}

Config file locations:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json

Add to your VS Code settings.json:

{
  "mcp.servers": {
    "envibe": {
      "command": "npx",
      "args": ["envibe-mcp"]
    }
  }
}

Add to Cursor MCP settings:

{
  "mcpServers": {
    "envibe": {
      "command": "npx",
      "args": ["envibe-mcp"]
    }
  }
}

Add to ~/.codeium/windsurf/mcp_config.json:

{
  "mcpServers": {
    "envibe": {
      "command": "npx",
      "args": ["envibe-mcp"]
    }
  }
}

# Install globally
npm install -g envibe

# Run setup
envibe setup

MCP Tools

| Tool | Description | |------|-------------| | env_list | List visible variables with access levels | | env_get | Get a variable's value (respects permissions) | | env_set | Set a variable (only full access) | | env_describe | Get detailed info including format and example | | env_check_required | Check which required variables are missing |

v0.2.0 Features

• Better error messages - When access is denied, get helpful guidance
• Format hints - Know what format a variable should be (url, key, number, etc.)
• Required var checking - Use env_check_required to guide users through setup

Why envibe?

| Approach | Problem | |----------|---------| | dotenvx | Encrypts files, but AI still needs the decryption key | | permissions.deny | Blocks all .env access—no granular control | | Just ignore .env | AI can't run or test code that needs env vars | | envibe | Per-variable access control. AI sees what you allow. |

File Structure

your-project/
├── .env                  # Real secrets (gitignored)
├── .env.example          # Template for devs (committed)
├── .env.manifest.yaml    # Access rules (committed)
├── .env.ai               # AI-safe view (gitignored)
└── .claude/
    └── settings.json     # Claude Code config (committed)

Security

• .env files are gitignored and blocked from AI
• .env.manifest.yaml contains rules only, not values—safe to commit
• .env.ai is regenerated from .env + manifest—gitignore it
• Unknown variables default to placeholder (fail-safe)
• Bash workarounds blocked (cat .env, head .env, etc.)

License

MIT

Built for the AI coding era. Stop leaking secrets.