This is a privacy protection layer that sits between Claude and your data, scanning for personally identifiable information and secrets before they hit the model. The OCR merging logic suggests it can handle both text and image inputs, identifying patterns like social security numbers, API keys, credit cards, and other sensitive data across different formats. You'd use this when working with customer data, support tickets, or any workflow where you need to ensure PII doesn't leak into prompts or get logged. It runs as a gateway rather than a direct integration, meaning it intercepts and sanitizes content in transit. Useful for compliance heavy environments where you want Claude's capabilities without the data exposure risk.
claude mcp add --transport sse io.github.evozim-zero-leak https://zero-leak-mcp.vercel.app/api/mcp