AgentGuard

AI agents are going autonomous. Claude Code writes 500 files overnight. Codex opens PRs. Marketing bots publish content. This week on HN: an AI agent published a hit piece on someone (2,021 points). Another opened a PR and wrote a blog post shaming the maintainer who closed it (904 points). Your agent could be doing something catastrophic right now. You'd never know until it's too late.

Middleware for AI agent observability. See every action your agents take in real-time. Set guardrails: no external API calls, no file deletions, no public posts without review. Get alerts when agents attempt risky actions. One-click kill switch. Works with Claude Code, Codex, custom agents. Because 'just trust the AI' stopped being a strategy.