Masquerade MCP – the privacy firewall for Claude

5 hours ago 1

mcp_demo.mp4

🤐 Problem: Tools like Claude or GPT are incredibly powerful, but they require raw input. If you're dealing with contracts, medical records, or internal documents, that's risky.

🛡️ Solution: Masquerade acts as a privacy firewall for your files. Just paste in the file path to a PDF, and Masquerade will:

Automatically detect sensitive data (names, emails, dates, entities)
Redact the sensitive data
Let you preview before sending to an LLM

User Input: The user asks Claude to redact a PDF by providing its file path.
PDF Processing: MCP reads the PDF and converts it to text.
Sensitive Data Detection: The text is sent to Tinfoil (an isolated AI platform using Llama 3.3 70B) to identify sensitive data.
Redaction: MCP removes the sensitive data and creates a redacted PDF.
Summary Return: MCP sends Claude a summary with:
- Masked versions of the sensitive data
- Redaction counts per page
- The path to the redacted file
Quering PDF with Claude: The user can upload the redacted PDF to Claude if satisfied with the redactions.

mcp_setup.mp4

Install Claude desktop
Get Tinfoil API key (create account and API key)
Configure environment

Option 1: Automated

curl -O https://raw.githubusercontent.com/postralai/masquerade/main/setup.sh && bash setup.sh

Restart Claude desktop app (if successfully configured)

Option 2: Manual (click to expand)

Create a virtual environment with Python ">=3.10, <=3.12"

python3.12 -m venv pdfmcp source pdfmcp/bin/activate python --version

Install this repo with the command below

pip install git+https://github.com/postralai/masquerade@main

Automate the Claude config setup (and skip the next steps)

python -m masquerade.configure_claude

Get Python path: which python
Get MCP file path: python -c "import masquerade as m; print(f'{m.__path__[0]}/mcp_pdf_redaction.py')"
Add (1) Python path, (2) MCP file path, and (3) Tinfoil API key to the JSON below and add that to claude_desktop_config.json. Instructions to find the config file are in the image below.
Restart Claude

{ "mcpServers": { "pdf-redaction": { "command": "/path/to/python", // Run `which python` "args": ["/path/to/mcp_pdf_redaction.py"], // Run `python -c "import masquerade as m; print(f'{m.__path__[0]}/mcp_pdf_redaction.py')"` "env": { "TINFOIL_API_KEY": "your_api_key" // Create Tinfoil account and paste API key } } } }

Prompt Claude: "Redact sensitive information from this PDF: /path/to/filename.pdf"
MCP opens the redacted and highlighted PDF files
Manually upload the redacted PDF to Claude for quering

Note: Don't upload the original PDF to Claude, only the file path.

If you have a fix or improvement, feel free to submit a pull request. For significant changes, please open an issue first to discuss your proposed updates.

See the Developer section for the installation of this repository.

This project is licensed under the MIT License - see the LICENSE file for details.

For questions and common issues, please see the FAQ section or open an issue on GitHub.

Read Entire Article