Skip to content

Privacy Filter

Overview

The privacy filter automatically strips sensitive data (API keys, secrets, tokens, private tags) from memory entries before storage.

Supported Patterns

  • OpenAI API keys (sk-*)
  • Anthropic API keys (sk-ant-*)
  • GitHub tokens (ghp_, gho_, ghs_, ghr_)
  • Slack tokens (xox*)
  • AWS keys (AKIA*)
  • Google API keys (AIza*)
  • Stripe keys (sk_live_, pk_live_, sk_test_)
  • Bearer tokens
  • Private/secret tags

Usage

The privacy filter is automatically applied in think and memory_remember before persistence.

from mcp_server.utils.privacy import strip_secrets, has_secrets

Strip secrets

clean_text = strip_secrets(text_with_secrets)

Check for secrets

if has_secrets(user_input): logger.warning('Input contains potential secrets')

Testing

pytest tests/test_privacy.py -v