@blindfold/mcp-server

MCP server for Blindfold — protect sensitive data in AI conversations. Detect, tokenize, mask, redact, hash, encrypt, or synthesize PII directly from Claude Desktop, Claude Code, Cursor, or any MCP-compatible client.

Setup

Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "blindfold": {
      "command": "npx",
      "args": ["-y", "@blindfold/mcp-server"],
      "env": {
        "BLINDFOLD_API_KEY": "your_api_key"
      }
    }
  }
}

Claude Code

Add to your project's .mcp.json:

{
  "mcpServers": {
    "blindfold": {
      "command": "npx",
      "args": ["-y", "@blindfold/mcp-server"],
      "env": {
        "BLINDFOLD_API_KEY": "your_api_key"
      }
    }
  }
}

Cursor

Add via Settings > MCP Servers, or in .cursor/mcp.json:

{
  "mcpServers": {
    "blindfold": {
      "command": "npx",
      "args": ["-y", "@blindfold/mcp-server"],
      "env": {
        "BLINDFOLD_API_KEY": "your_api_key"
      }
    }
  }
}

Get an API Key

1. Sign up at app.blindfold.dev
2. Go to API Keys and create a new key
3. Copy the key into your config

Available Tools

| Tool | Description | |------|-------------| | blindfold_detect | Detect PII without modifying text | | blindfold_tokenize | Replace PII with reversible tokens | | blindfold_detokenize | Restore original values from tokens | | blindfold_mask | Partially hide sensitive data | | blindfold_redact | Permanently remove PII | | blindfold_synthesize | Replace with realistic fake data | | blindfold_hash | One-way hash PII | | blindfold_encrypt | Encrypt PII with a password | | blindfold_discover | Analyze samples for PII types |

Usage Examples

Once configured, you can ask Claude to use Blindfold tools naturally:

Tokenize before processing:

"Tokenize this patient record before summarizing: John Doe, SSN 123-45-6789, diagnosed with diabetes"

Redact logs:

"Redact all PII from these server logs before analyzing them"

Analyze data for PII:

"Discover what types of PII are in these customer feedback samples"

Synthesize test data:

"Replace real customer data in this CSV with synthetic data"

Recommended Instructions

For automatic PII protection, add to your project's CLAUDE.md or system prompt:

Always use blindfold_tokenize before processing any text that contains
personal information (names, emails, phone numbers, addresses, SSNs,
medical data). After generating a response, use blindfold_detokenize
to restore original values.

Detection Policies

All tools accept an optional policy parameter:

| Policy | Use Case | |--------|----------| | basic | Common PII (names, emails, phones) | | strict | Maximum detection, all entity types | | gdpr_eu | EU GDPR-relevant entities | | hipaa_us | US healthcare (PHI, SSN, insurance) | | pci_dss | Payment card data (credit cards, IBANs) |

Environment Variables

| Variable | Required | Default | Description | |----------|----------|---------|-------------| | BLINDFOLD_API_KEY | Yes | — | Your Blindfold API key | | BLINDFOLD_BASE_URL | No | https://api.blindfold.dev | API base URL |

Security

• Your API key is stored locally in your config file and never sent to the AI model
• The MCP server runs as a local process on your machine
• All API calls use HTTPS
• The AI model only sees tool names, parameters, and results

Development

npm install
npm run build
npm run dev   # watch mode

License

MIT