VibeGit: Keeping AI Coding History Off Your Git History
When you code with AI assistants like Claude Code or Cursor, the back-and-forth between you and the AI can generate dozens of micro-iterations before you're ready to commit. Traditional Git commits are too coarse for this workflow — you don't want 50 "WIP" commits cluttering your history.
That's where MemoV comes in.
The Problem with AI Coding Sessions
Consider a typical AI coding session:
- You ask the AI to implement a feature
- The AI writes initial code
- You refine the prompt, the AI adjusts
- You spot a bug, the AI fixes it
- You ask for tests, the AI adds them
- Finally, you commit
Each step involves code changes, but you only want step 6 in your Git history. Yet losing the earlier steps means losing valuable context: What prompts led to this solution? What alternatives were tried?
MemoV's Shadow Timeline
MemoV creates a parallel timeline in the
directory that captures every AI interaction:code.mem
codeyour-project/ ├── .git/ # Your clean commit history ├── .mem/ # MemoV's shadow timeline │ ├── memov.git/ # Bare Git repository for snapshots │ ├── branches.json │ └── jump.json └── src/
The shadow timeline stores:
- Prompts — What you asked the AI
- Responses — How the AI explained its changes
- Code diffs — Exactly what changed at each step
Key Benefits
1. Clean Git History
Your main repository only contains intentional, well-described commits. AI experimentation stays separate.
2. Full Traceability
Need to understand why code looks a certain way? Check the MemoV history to see the prompts that shaped it.
3. Easy Rollback
Made a wrong turn with the AI? Jump back to any snapshot:
bashmem history # See available snapshots mem jump abc123 # Restore that state
Your Git working directory updates, but no Git commits are created.
4. Branch Experiments
Try different approaches on different MemoV branches:
bashmem snap --prompt "Approach A: use recursion" # ... experiment ... mem jump earlier-snapshot mem snap --prompt "Approach B: use iteration"
5. Privacy by Default
Everything stays local. Add patterns to
to exclude sensitive files from being tracked.code.memignore
How Teams Use MemoV
- Solo developers — Keep personal experimentation history for learning and debugging
- Code review — Share the MemoV timeline to explain how you arrived at a solution
- Onboarding — New team members can see how features were built, not just the final code
- Auditing — Track exactly which AI prompts led to security-sensitive code
Getting Started
- Install MemoV:
bashcurl -LsSf https://astral.sh/uv/install.sh | sh
- Add to your AI coding tool:
bashclaude mcp add mem-mcp --scope project -- uvx --from git+https://github.com/memovai/memov.git mem-mcp-launcher stdio $(pwd)
- Start coding — MemoV captures automatically!
Read the full installation guide or explore core concepts to learn more.