MCP Connector for Claude. Redact PII & Catch Leaked Keys

What this connector does

Privacy Prompt AI Protector is the MCP (Model Context Protocol) companion to PrivacyPromptAI. It gives Claude two tools it can call automatically during a chat, so sensitive text never has to leave the conversation unprotected:

Redact PII from any pasted text or file content before it's summarized, forwarded, or logged.
Scan a prompt for exposed credentials — API keys, tokens, passwords — before it's sent onward.

Both tools run statelessly: each request is processed in server memory only, for the duration of that single call. Nothing submitted is written to disk, logged, stored in a database, or shared with any third party.

Endpoint & connection details

Server URL	`https://mcp.privacypromptai.com/mcp.php`
Transport	Streamable HTTP (stateless)
Authentication	None — no account or login required
Capabilities	Read-only (both tools only read the text you provide and return a result)

Tools

1. `clean_pii_from_text`

Redacts common PII from a block of text:

Detected	Replaced with
Email addresses	`[REDACTED_EMAIL]`
Phone numbers	`[REDACTED_PHONE]`
Credit card numbers (15-digit Amex & 16-digit Visa/Mastercard/Discover)	`[REDACTED_CARD_NUM]`
IPv4 addresses	`[REDACTED_IP]`

Input: "Contact John at [email protected] or +1 (555) 123-4567."
Output: "Contact John at [REDACTED_EMAIL] or [REDACTED_PHONE]."

2. `evaluate_prompt_safety`

Audits a prompt for signs of exposed credentials before it's sent to an external model: OpenAI-style keys (sk-...), Anthropic-style keys (sk-ant-...), AWS access key IDs (AKIA...), bearer tokens/JWTs, PEM private key headers, and generic api_key=/secret=/password=/token= assignments.

Input: "here is my key sk-abcdefghijklmnopqrstuvwx"
Output: "⚠ Potential credential exposure detected… found an OpenAI-style secret key (sk-...)."

Set up in Claude Desktop

Add the following to your claude_desktop_config.json (Settings → Developer → Edit Config), then fully restart Claude Desktop:

{
  "mcpServers": {
    "privacy-prompt-ai-protector": {
      "url": "https://mcp.privacypromptai.com/mcp.php"
    }
  }
}

Try pasting a paragraph with a fake email address or phone number and asking Claude to sanitize it — it should call clean_pii_from_text automatically and return the redacted version.

Data handling

This connector follows the same zero-retention principle as the rest of PrivacyPromptAI. Full details are in our Privacy Policy, under "MCP Connector (AI Assistant Integration)."

Questions or issues? Contact [email protected].

Privacy Prompt AI Protector: MCP Connector for Claude