We build tools that show you what your AI is actually doing.

AI coding assistants like Claude Code can read any file in your project, execute arbitrary commands, and make network requests — all without an audit trail. Developers trust these tools because they're productive. But trust without visibility is a liability.

We saw this gap growing every day. Millions of developers running AI tools that operate unsupervised inside their most sensitive codebases. No logs. No rules. No way to know what actually happened after a session.

Harpax is our first product. It monitors every action your AI coding assistant takes and gives you a complete, searchable record of what happened. It can block dangerous actions in real time or run passively and just log. It's open source, self-hosted, and all data stays local.

It works today. You can install it in under two minutes.

We have two more products in development. We're not ready to talk about them yet — we'll ship when they're real, not when they're interesting to pitch.

If you're interested in AI governance tooling, or if you're building in this space, we'd like to hear from you.