A self-contained snapshot and context tool for your AI development workflow.
We've all been there. You're deep in an LLM project or machine learning development, juggling code, prompts, and a chaotic mess of trial-and-error. Git feels too heavy for these quick, experimental loops. Your terminal history is a graveyard of forgotten commands, and your text editor is a minefield of unsaved changes.
That's why I made DevCat. It's a micro-version control system built for how we actually work. No branches, no commits, just instant, clean snapshots of your project. It's a lifesaver when you need to quickly save a good state, revert to a previous version, or grab a full code context to feed back to an AI model. DevCat gets out of your way and gives you a fighting chance in the wild world of AI development.
- Snapshot & Revert: Take a snapshot of your project with a simple message. This is essential for quickly saving states during AI model tuning or prompt engineering. Later, you can jump back to any of those snapshots instantly if things go sideways.
- Context Generation: Ever need to give an LLM a full view of a module or your entire project? DevCat concatenates all your relevant code into a single stream, ready for copy/pasting or piping. This is perfect for providing AI models with the context they need.
- Diffing: Compare snapshots to see what changed, or check the difference between your last saved state and your current work.
- Trace Extraction: Pipe in a stack trace from your application, and DevCat will fetch the exact code snippets from each file reference, giving you instant debugging context.
- Auto-Watch: Let DevCat handle the saves for you. It can watch your files and automatically create snapshots when it detects changes.
- Zero-Config: Just download it and run it. No complex setup or remote servers. It just works.
DevCat is a Rust binary.
git clone https://github.com/gni/devcat.git
cd devcat
cargo install --path .cargo install devcatHere’s a quick rundown of the essential commands.
Save a new snapshot of the current project state.
# Save your current progress
devcat save "refactored the prompt engineering logic"
# Save a snapshot but exclude the 'data' folder
devcat save "experimenting with new data loaders" --exclude 'data/'Show a history of all saved snapshots.
devcat logOutput:
ID TIMESTAMP MESSAGE
--- ---------------------- --------------------------------------------------
1 2025-08-04 17:01:44 Initial commit
2 2025-08-04 17:05:21 Refactored the core loop
3 2025-08-04 17:12:03 Implemented new AI logic
This is the default command. It concatenates all files in the current directory (or a specified path) into a single output. It's perfect for giving context to an LLM.
# Send all files in the current directory to your LLM
devcat | pbcopy
# Concatenate files from a specific module
devcat ./src/my_moduleConcatenate all files from a specific snapshot.
# Re-run a prompt with the state from snapshot 2
devcat --id 2 | some_llm_cli "debug this"Show the differences between two snapshots or between a snapshot and your current working directory.
# Compare the latest snapshot with your current work
devcat diffRevert the working directory to the state of a specific snapshot.
# Revert to a stable state
devcat revert 2Pipes in a stack trace and outputs the referenced code snippets.
# Get context for a Python stack trace
python my_script.py 2>&1 | devcat traceStarts a file watcher that automatically creates a new snapshot whenever files are changed.
# Don't think about it, just code. DevCat will handle the snapshots.
devcat watchFor now, DevCat is mostly plug-and-play. If you want to permanently exclude files or directories, you can create a .devcatrc file in the root of your project.
# Exclude the 'target' and 'node_modules' directories globally
exclude = ["target/", "node_modules/"]DevCat is an open-source project. If you have an idea, a bug report, or a feature request, feel free to open an issue or submit a pull request.
Lucian BLETAN - initialized project