Agent Guard

Agent Guard is an AI-agent security tool that scans your local development environment for signals that can get AI coding accounts banned, then fixes them. It checks 20+ risk signals across shared environment factors and six per-product profiles — Claude Code, Cursor, Codex, Gemini, Windsurf, and Hermes — flagging issues like proxy leaks, timezone-versus-IP mismatches, shell configuration residue (for example HTTPS_PROXY or ANTHROPIC_BASE_URL in .zshrc), config leaks, and plaintext MCP API keys. Scanning is free and runs entirely on your machine, producing a red/yellow/green report with a score and file-level detail on each finding. The Pro version adds one-click auto-fix for supported items — timezone correction, shell proxy cleanup, agent config cleanup, and disabling auto-timezone — and it backs up original files to ~/.agent-guard/backups/ before any change so you can roll back. A newer runtime security guard adds continuous monitoring with 40+ safety rules. Agent Guard runs on macOS (Apple Silicon and Intel) and Windows. The free scan is available now; Pro auto-fix pricing is not published on the site.

Best for developers using Claude Code, Cursor, or Codex who want to check their local setup for account-ban risk signals.