Agent Guard
Agent Guard：给 AI 编程助手做一次环境合规体检
Start new thread

Agent Guard - Local security & compliance scanner for AI agents

by•
Agent Guard now adds runtime security hooks to its local AI-agent scanner. Detect 20+ ban-risk signals—proxy leaks, timezone/location mismatches, shell config residue, and plaintext MCP keys—across Claude, Cursor, Codex, Gemini, Windsurf, and Hermes. Free scan gives you a risk report with manual fixes; Pro adds one-click auto-fix with automatic backups. macOS & Windows.

Add a comment

Replies

Best
Maker
📌
I started Agent Guard after seeing too many Claude, Cursor, and Codex accounts get banned not because of what users did, but because of tiny environment leaks—a proxy left in shell config, a timezone mismatch, or an MCP key sitting in plaintext. This launch adds runtime security hooks on top of the local environment scan, so risky agent actions (like destructive commands or unexpected network calls) get intercepted before anything bad happens. Everything stays local on your machine, with no data sent anywhere. Free scan gives you a full risk report with manual fixes; Pro adds one-click auto-fix with automatic backups. I'd love your feedback on what other risks we should catch next.
💎 Pixel perfection

The one-click repair flow is genuinely clever, turning what could be a deep technical headache into something anyone can actually use.

Detecting plaintext MCP keys and proxy leaks across Claude/Cursor/Codex is a real gap — do you also flag credentials for third-party gateway endpoints those agents point at? Many of our users route Claude Code through a custom base URL, so key hygiene there matters a lot.

Maker

 Any logs or information scanned or collected by this application are stored locally on the user's hard drive; no data is uploaded to our servers. We only use event tracking to record which buttons users click, helping us continuously optimize the product.