PDF Redaction MCP

PDF Redaction exposes its PII detection and redaction engine as an MCP (Model Context Protocol) server. This lets any MCP-compatible AI client — Claude Desktop, Claude Code, Cursor, and others — call PDF redaction tools directly as part of an agent's toolset, without writing any REST integration code.

Endpoint

https://api.pdf-redaction.com/mcp/

The server uses the Streamable HTTP transport and is served alongside the existing REST API.

Authentication

This server uses OAuth 2.1. The first time your client calls a tool, it opens your browser to log in to your PDF Redaction account (or create one — magic link, Google, Microsoft, and LinkedIn are all supported) — no API key required. Claude Desktop, Claude Code, Cursor, and other OAuth-capable MCP clients handle this automatically as part of connecting.

Once authorized, tool calls run against your account's usage and credits, the same as the REST API. If credits run out, calls return an MCP tool error describing the problem (e.g. Credits exhausted).

Free tier limits

• 10 pages per request
• 100 free pages per month
• 5 requests per minute

Configuring your client

Claude Desktop / Claude Code / Claude.ai

Click Connect to Claude above, or add it manually:

1. In Claude, go to Settings → Connectors → Add custom connector.
2. Enter https://api.pdf-redaction.com/mcp/ as the URL.
3. Approve the login prompt that opens in your browser.

For Claude Code, you can instead run:

claude mcp add --transport http pdf-redaction https://api.pdf-redaction.com/mcp/

Cursor

Add the server to .cursor/mcp.json:

{
  "mcpServers": {
    "pdf-redaction": {
      "url": "https://api.pdf-redaction.com/mcp/"
    }
  }
}

Cursor opens a browser login prompt the first time you use a tool from this server.

ChatGPT

ChatGPT doesn't support one-click connector installs yet, so add the server manually:

1. In ChatGPT, enable Developer mode (Settings → Apps & Connectors → Advanced settings).
2. Go to Connectors → Create, and enter https://api.pdf-redaction.com/mcp/ as the server URL.
3. Choose OAuth authentication and approve the login prompt that opens in your browser.

Available tools

anonymize_pdf

Detects PII in a PDF and returns a redacted copy of the document.

Returns: { detected_pii, processing_time, download_url } — download_url is a secure, time-limited link to download the redacted PDF. If include_pdf_base64 is true, the response also includes pdf, the base64-encoded redacted document.

anonymize_pdf_from_upload

Anonymizes a PDF that was uploaded to the server beforehand, instead of passing it inline as base64. Useful for large files or clients with built-in file-upload support (e.g. Claude Desktop, claude.ai).

1. Upload the PDF via the client's file upload UI, or call the store_files tool directly.
2. Call anonymize_pdf_from_upload with the uploaded file's file_name.

Returns: { detected_pii, processing_time, download_uri, download_url } — download_uri is an MCP resource URI (redaction://...) and download_url a secure, time-limited https:// link, both pointing at the redacted PDF. The result is also saved as <file_name>-redacted.pdf, retrievable later via the list_files/read_file tools. If include_pdf_base64 is true, the response also includes pdf.

detect_pii_pdf

Detects PII in a PDF without redacting it.

Returns: { detected_pii, processing_time }

anonymize_pdf_custom

Redacts a PDF using a free-form natural-language instruction instead of a predefined tag list.

Returns: { detected_pii, processing_time, download_url }, and optionally pdf if include_pdf_base64 is true.

Supported tags

DATE, PERSON_NAME, ORGANIZATION, LOCATION, EMAIL, PHONE, ID, ACCOUNT, ZIP_CODE, ADDRESS, IP, URL, SSN, DRIVER_LICENSE, PASSPORT, PASSWORD, AGE, CREDIT_CARD, MONEY_AMOUNT, SIGNATURE, QR_CODE, FACE

Supported OCR languages

eng, spa, fra, deu, ita, por, rus