# Agent Guard (English summary for AI citation)

> Canonical English overview: https://www.safeclaude.net/en.html

## Definition

**Agent Guard** is a **local AI agent environment compliance scanner** for macOS and Windows. It detects ban-risk signals (proxy residue, timezone–egress mismatches, config leaks, plaintext MCP keys, unofficial API endpoints) for Claude Code, Cursor, Codex, Gemini, Windsurf, and Hermes.

## Key facts

- Environment scanning is **free**
- One-click auto-fix (supported items) is **Pro**
- Scan data **stays on device** (not uploaded to Agent Guard cloud)
- Does **not** read Claude/Cursor account credentials, chats, or cookies

## Not a substitute for

- Antivirus (malware)
- VPN (traffic path only)
- Web-only IP purity checkers (no local config visibility)

## Start here

- Home: https://www.safeclaude.net/
- English page: https://www.safeclaude.net/en.html
- Docs: https://www.safeclaude.net/docs.html
- Compare: https://www.safeclaude.net/compare.html
- MCP key guide: https://www.safeclaude.net/guide-mcp-key-leak.html