A macOS utility for selecting, combining, and copying the contents of files — ideal for use with LLMs and generative AI tools. FileKitty lets you grab context from multiple files with one click and keeps a full history of your selections.
LLM tools like Cursor, GitHub Copilot Chat, and Claude Code are powerful — but behind the scenes, they often make API completion requests using bloated, overstuffed prompts. These include entire repos, unrelated files, and boilerplate-heavy context, which can degrade the quality of even state-of-the-art models.
FileKitty gives you control over the prompt. It’s a lightweight tool for assembling precise, readable context that pairs perfectly with interactive (chat-based) LLM workflows — whether you're using ChatGPT, Claude, Gemini, or any number of local models.
- For coding: Select only the files, classes, or functions you're working on — and get direct, targeted suggestions instead of vague completions.
- For documentation and planning: Combine real code with Slack threads, meeting notes, or config files to produce accurate explanations, writeups, and planning docs.
- For debugging and support: Isolate logs, config diffs, and source snippets into a focused prompt — no need to paste in your whole project.
Unlike IDE-native assistants that guess behind the scenes, FileKitty lets you see and shape exactly what the model sees. You get cleaner inputs, smarter outputs, and fewer wasted tokens.
What it lacks in IDE integration, it makes up for in speed, precision, and higher-quality problem solving.
- Gather code snippets for LLM prompts (ChatGPT, Claude, Gemini, Copilot, etc.)
- Provide precise multi-file context for generative AI tools
- Combine logs, configs, or structured docs for inspection
- Track and revisit prior file selections and outputs
brew install banagale/filekitty/filekitty- From Terminal:
filekitty
- Or via Finder:
open /opt/homebrew/opt/filekitty/FileKitty.app
To access via Spotlight or Launchpad, copy the app to /Applications:
ditto /opt/homebrew/opt/filekitty/FileKitty.app /Applications/FileKitty.appNote: Using ditto rather than cp preserves the application bundle's icon and metadata.
Install Poetry and build locally:
git clone https://github.com/banagale/FileKitty.git
cd FileKitty
poetry install
poetry run python setup.py py2appThe app will be created in ./dist/. Copy it to /Applications for full integration.
Manual builds are useful for development or Linux/Windows adaptation.
Here’s what FileKitty produces when you drop a folder or select multiple files. The result is a clean, timestamped view of file contents, optionally preceded by a project folder tree.
Below, we’ve selected three Python files from a project:
The output looks like this inside FileKitty’s preview pane:
# Folder Tree of ~/code/…/FileKitty/src/filekitty
```text
filekitty/
├── core/
│ ├── qt_imports.py
├── resources/
├── ui/
│ ├── __pycache__/
│ ├── components/
│ ├── dialogs.py
│ ├── main_window.py
│ ├── qt_widgets.py
│ ├── text_output_area.py
│ └── tree_settings_dialog.py
├── __main__.py
├── app_logic.py
├── constants.py
└── ...
```
# \~/code/…/filekitty/**main**.py
**Last modified: May 20, 2025 5:07 PM**
```python
from filekitty.app_logic import main
if __name__ == "__main__":
main()
```
# \~/code/…/filekitty/constants.py
**Last modified: Jun 3, 2025 10:26 AM**
```python
from pathlib import Path
ICON_PATH = str((Path(__file__).parent / "resources/icon/FileKitty-icon.png").resolve())
SETTINGS_FILE_IGNORE_KEY = "mainOutputIgnoreRegex"
...
```
# ~/code/…/filekitty/ui/main_window.py
**Last modified: Jun 5, 2025 12:54 PM**
```python
class FilePicker(QWidget):
def __init__(self, initial_files=None):
...- Open the app
- Select Files or drag-and-drop files from Finder, PyCharm, etc.
- Combined contents will appear, grouped in Markdown code blocks.
- Click Copy to Clipboard and paste into your LLM or chat.
- Back/Forward buttons let you navigate prior selections
- Changes to file contents are detected and marked as
(Modified)or(Missing Files)
- When
.pyfiles are selected, use Select Classes/Functions to target specific symbols and relevant imports.
- Click Refresh to reload the current selection, useful after editing source files.
Access via FileKitty → Preferences (Cmd+,):
- Default Select Directory – sets initial folder for file dialog
- History Location – controls where snapshot state is stored
git clone https://github.com/banagale/FileKitty.git
cd FileKitty
poetry install-
The app can be manually built using:
poetry run python setup.py py2app
The resulting
.appbundle appears in./dist/. -
Run the app directly with:
poetry run filekitty
Before building or releasing, make sure you have:
- macOS (required for app build)
- Poetry installed
- Homebrew - only required for formula validation using filekitty-validate. Must be available in your PATH.
FileKitty ships two developer tools for publishing new versions:
| Command | What it does |
|---|---|
poetry run filekitty-release |
Interactive release assistant: verifies version consistency, builds the app, zips it, generates a .sha256, shows a changelog, tags the release, and outputs GitHub/Homebrew update instructions. |
poetry run filekitty-validate |
Post-release checker: ensures the pushed tag exists, the ZIP file matches the hash, and the Homebrew formula matches the release metadata. |
Tip: Add
--dry-runto simulate the release without making changes.
Run linting and formatting tools:
make lint
make formatOr directly with Poetry:
poetry run ruff check .
poetry run ruff format .Set up automatic checks before commits:
pre-commit install
pre-commit run --all-files- GitHub Actions validate builds and style checks on each push.
- See
.github/workflows/for automation details.
MIT License © Rob Banagale