About ProxyGuard

Why ProxyGuard Exists

AI coding assistants like Claude and Cursor send code context to cloud APIs. That context can accidentally include secrets like API keys, database credentials, and access tokens. Most developers don't realize their environment variables are being sent until it's too late. ProxyGuard sits between your tools and the internet, catching secrets before they leave your Mac.

How It Works

ProxyGuard runs a local proxy on port 8080. When you route your AI tools through it, every outgoing request is scanned for patterns that match known secret formats like AWS keys, GitHub tokens, Stripe keys, and database connection strings. Matches are redacted in real time. Clean requests pass through untouched. The entire process adds less than 5 milliseconds of latency.

ProxyGuard currently detects over 18 secret patterns: AWS access and secret keys, GitHub personal access tokens, Stripe API and webhook keys, OpenAI and Anthropic API keys, Slack bot tokens, PostgreSQL, MySQL, and MongoDB connection strings, JWT secrets, and SSH private key material. New patterns are added with each software update.

Built in Pure Swift

No Electron. No JavaScript runtime. No background services phoning home. ProxyGuard is a native macOS app written entirely in Swift. It lives in your menu bar, uses minimal system resources, and works offline after a one-time license activation. Your security tool should not itself be a security risk.

Who Makes ProxyGuard

ProxyGuard is built by Alejandro Roman, a software developer focused on developer tooling and security. It started as a personal tool after noticing that Claude Code was occasionally picking up secrets from environment files during context gathering. Rather than relying on manual vigilance, a local proxy that redacts automatically felt like the right solution. ProxyGuard is that tool, made available to any developer who works with AI coding assistants on macOS.

For questions, feature requests, or support, reach out at me@alejandroman.dev.