Introducing mrq: Version Control for AI-Assisted Development

Over the past year, we’ve been watching how developers actually use AI coding assistants like Cursor, Claude, and GitHub Copilot. One thing became clear: the way these tools work doesn’t match the assumptions built into git.

Git is designed around deliberate, discrete commits with meaningful messages, which works wonderfully for that use case. But when you’re iterating rapidly with an AI assistant (trying different approaches, making sweeping changes, exploring solutions) the commit model adds friction. Most developers either stop committing during AI-assisted sessions or create a mess of checkpoint commits.

We built mrq to address this gap.

What mrq Does

mrq runs in the background and automatically captures snapshots of your codebase as you work. There’s no staging, no commit messages, no manual intervention required. When meaningful changes happen, they’re preserved automatically.

$ mrq watch --daemon
✓ Workspace created: my-project  
✓ Watch daemon started (PID: 48291)

Once it’s running, you can continue working normally. The daemon handles everything in the background.

When something goes wrong (and with AI assistants making rapid changes, things occasionally do) you can restore any previous state:

$ mrq restore abc123
✓ Restored to snapshot abc123
✓ Backup created at .mrq-backup-1703001234

Your current state is automatically backed up before the restore, so you can always change your mind.

How It Fits with Git

mrq works alongside git rather than replacing it. You still use git for collaboration, code review, pull requests, and deployment pipelines. mrq adds a safety net during the exploratory phase of development, particularly the messy, iterative work that’s common when using AI assistants.

Think of it as local backup that understands your development workflow. Once you’re happy with where your code landed, you commit to git as usual.

What We’re Launching

Today we’re opening mrq to beta users. The core functionality is ready:

Automatic file system monitoring and snapshot creation
Instant restoration to any previous state
AI-generated summaries of what changed in each snapshot
A dashboard for browsing your workspace history
End-to-end encryption so your code stays private

We’re offering a free tier that’s sufficient for most individual developers, with paid plans for heavier usage.

Getting Started

Installation takes about a minute:

npm install -g mrq-cli@latest
mrq login
mrq watch

That’s it. Your file changes are now being captured automatically.

If you’ve ever lost work during an AI-assisted coding session, or found yourself reluctant to try experiments because you weren’t sure you could get back to a working state, we think you’ll find mrq useful.

We’d love to hear your feedback as you try it out.