Skip to main content
The Blindfold MCP (Model Context Protocol) Server lets AI assistants like Claude Desktop, Cursor, and Windsurf detect and protect PII directly through tool calls. No code required — the AI assistant handles everything.

​
Installation

npm install -g @blindfold/mcp-server

​
Configuration

The MCP server is configured via environment variables:
VariableRequiredDescription
BLINDFOLD_API_KEYYesYour Blindfold API key
BLINDFOLD_REGIONNoRegion for data residency: eu or us
BLINDFOLD_BASE_URLNoCustom API base URL (overrides region)
See Regions for data residency details.

​
Setup with Claude Desktop

Add to your Claude Desktop configuration file:
Edit ~/Library/Application Support/Claude/claude_desktop_config.json:
{
  "mcpServers": {
    "blindfold": {
      "command": "npx",
      "args": ["-y", "@blindfold/mcp-server"],
      "env": {
        "BLINDFOLD_API_KEY": "your-api-key-here",
        "BLINDFOLD_REGION": "eu"
      }
    }
  }
}
Restart Claude Desktop after saving.

​
Setup with Cursor

Add to your Cursor MCP settings (.cursor/mcp.json in your project or global settings): 
{
  "mcpServers": {
    "blindfold": {
      "command": "npx",
      "args": ["-y", "@blindfold/mcp-server"],
      "env": {
        "BLINDFOLD_API_KEY": "your-api-key-here",
        "BLINDFOLD_REGION": "eu"
      }
    }
  }
}

​
Setup with Claude Code

Add the MCP server to Claude Code: 
claude mcp add blindfold -- npx -y @blindfold/mcp-server
Set the required environment variable: 
export BLINDFOLD_API_KEY=your-api-key-here
export BLINDFOLD_REGION=eu

​
Available Tools

Once connected, the AI assistant has access to these tools:

​
blindfold_detect

Detect PII in text without modifying it. Returns detected entities with types, positions, and confidence scores. Parameters:
  • text (string) — Text to analyze, OR
  • texts (string[]) — Array of texts for batch processing (max 100)
  • policy (string, optional) — Detection policy
  • entities (string[], optional) — Specific entity types to detect
  • score_threshold (number, optional) — Minimum confidence (0.0-1.0)

​
blindfold_tokenize

Replace PII with reversible tokens. Returns tokenized text and a mapping to restore values later. Parameters:
  • text (string) — Text to tokenize, OR
  • texts (string[]) — Array of texts for batch processing (max 100)
  • policy (string, optional) — Detection policy
  • entities (string[], optional) — Specific entity types to detect
  • score_threshold (number, optional) — Minimum confidence (0.0-1.0)

​
blindfold_detokenize

Restore original values from tokenized text using a mapping. Parameters:
  • text (string) — Text containing tokens
  • mapping (object) — Token-to-original mapping from a previous tokenize call

​
blindfold_mask

Partially hide PII with masking characters. Parameters:
  • text (string) — Text to mask, OR
  • texts (string[]) — Array of texts for batch processing (max 100)
  • policy (string, optional) — Detection policy
  • masking_char (string, optional) — Masking character (default: *)
  • chars_to_show (number, optional) — Characters to leave visible
  • from_end (boolean, optional) — Show visible chars from end (default: true)

​
blindfold_redact

Permanently remove PII from text. Parameters:
  • text (string) — Text to redact, OR
  • texts (string[]) — Array of texts for batch processing (max 100)
  • policy (string, optional) — Detection policy
  • entities (string[], optional) — Specific entity types to redact
  • score_threshold (number, optional) — Minimum confidence (0.0-1.0)

​
blindfold_synthesize

Replace PII with realistic fake data. Supports 8 languages. Parameters:
  • text (string) — Text to synthesize, OR
  • texts (string[]) — Array of texts for batch processing (max 100)
  • policy (string, optional) — Detection policy
  • language (string, optional) — Language: en, cs, de, fr, es, it, pl, sk

​
blindfold_hash

Replace PII with deterministic one-way hashes. Parameters:
  • text (string) — Text to hash, OR
  • texts (string[]) — Array of texts for batch processing (max 100)
  • policy (string, optional) — Detection policy
  • hash_type (string, optional) — Algorithm: md5, sha1, sha256 (default)
  • hash_length (number, optional) — Truncate hash length

​
blindfold_encrypt

Encrypt PII with AES encryption. Parameters:
  • text (string) — Text to encrypt, OR
  • texts (string[]) — Array of texts for batch processing (max 100)
  • policy (string, optional) — Detection policy
  • encryption_key (string, optional) — Encryption password

​
blindfold_discover

Analyze text samples to discover what types of PII they contain. Parameters:
  • samples (string[]) — Array of text samples to analyze
  • threshold (number, optional) — Detection confidence threshold

​
Usage Examples

Once the MCP server is connected, you can ask your AI assistant things like:
  • “Tokenize this text before sending it to the API: John Doe, john@example.com”
  • “Detect what PII is in this customer message”
  • “Redact all personal data from this document using GDPR policy”
  • “Mask the credit card numbers in this text”
  • “What types of PII are in these sample texts?”
The AI assistant will automatically call the appropriate Blindfold tool.

​
Troubleshooting

​
Server not starting

Make sure BLINDFOLD_API_KEY is set: 
echo $BLINDFOLD_API_KEY

​
Invalid region error

Valid regions are eu and us: 
export BLINDFOLD_REGION=eu

​
Testing manually

Run the server directly to check for errors: 
BLINDFOLD_API_KEY=your-key npx @blindfold/mcp-server
The server communicates via stdio, so you should see no output if it starts successfully.