Masquerade
Provides a privacy firewall for large language models, automatically detecting and redacting sensitive information from PDF documents before processing.
Acerca de
Large Language Models (LLMs) like Claude or GPT offer immense power but necessitate raw input, posing significant risks when handling sensitive documents such as contracts, medical records, or internal files. Masquerade addresses this by acting as a privacy firewall, intelligently detecting and redacting personal or confidential data (like names, emails, and entities) from PDFs. It allows users to preview the redacted document before interaction with an LLM, ensuring data privacy and compliance while leveraging AI capabilities. The system processes PDFs, uses an isolated AI for sensitive data identification, and then generates a summary including masked data and redaction details for the user and the LLM.
Características Principales
- Automatically detects sensitive data (names, emails, dates, entities)
- Redacts sensitive data from PDF files
- Allows preview of redacted documents before sending to LLMs
- 43 GitHub stars
- Integrates seamlessly with Claude desktop via Model Context Protocol (MCP)
- Generates summaries with masked sensitive data and redaction counts
Casos de Uso
- Anonymizing medical records for AI analysis
- Handling internal company documents securely with external LLM services
- Safely processing confidential contracts with LLMs
