Show HN: Git for AI Agents

What it is

Picture Git, but instead of tracking code changes, it tracks agent decisions. Every time an AI agent acts—deleting files, refactoring code, making API calls—re_gent creates a commit-like snapshot. You get a timeline of agent behavior with the reasoning attached. It's version control for the "why" behind agent actions, not just the "what."

Why it matters

If you're using AI agents for real work, you've hit this: the agent breaks something, and you have no idea when or why it happened. No git log for agent behavior. Re_gent gives you bisect-style debugging for agent sessions—rewind to find the exact moment things went wrong, see the agent's reasoning, restore previous states. Essential infrastructure as agents get more autonomous.

Key details

• •Open source project called re_gent, available on GitHub at regent-vcs/re_gent
• •Answers questions agents can't: "when did you delete this?", "why did you make that change?"
• •Supports rewinding and bisecting through agent sessions to debug behavior
• •Designed to work across sessions, not just within a single chat thread
• •Targets the gap between agent capabilities and developer workflow needs

Worth watching

0:46

Introducing Agent HQ: your mission control for AI agents

GitHub

GitHub's official introduction to Agent HQ provides authoritative insight into the platform's native tooling for managing and controlling AI agents at scale.

25:08

GHU: Building AI Agents with VS Code and GitHub

Microsoft Developer

Microsoft's technical walkthrough on building AI agents with VS Code and GitHub demonstrates practical integration patterns with industry-standard development tools.

6:58

How the GitHub Copilot coding agent works | GitHub Checkout

GitHub

GitHub's technical breakdown of how Copilot's coding agent works offers deep understanding of the architecture and decision-making processes behind AI-powered code generation.