FAQ
Common questions about tokens, privacy, the app's limits and troubleshooting.
Tokens and access
Why a token rather than signing in?
The app deliberately does not use OAuth. OAuth needs a registered client and a browser round trip, which does not fit "a single portable exe". A Personal Access Token is simpler and more transparent: you decide what it can do, and when to revoke it.
Does the app phone anywhere unexpected?
Only two places, both configurable in Settings:
- The GitHub API —
api.github.com - Whichever AI provider you chose
If you pick a local model, traffic stays between your machine and localhost.
What if my token leaks?
Delete it on >github.com/settings/tokens. It stops working immediately and the app stops using it.
Privacy
Where does my repository content go?
Only where you send it. When you ask the AI for something, the file tree and a few key files are sent to the model — but that only happens when you press the button, never automatically.
Does GitHub store my access?
Not remotely, in fact: your token is stored only on your own machine, encrypted. There is no server in the middle. This app talks to GitHub, not to anyone else.
What does the app store?
data/
settings.json appearance and AI settings
secrets.json token and API key, encrypted with DPAPI
That is all. No logs, no analytics, no crash reporting.
Can I keep the data folder on a USB stick?
Yes — that is exactly what it is designed for. The data folder stays next to
the executable and travels with it.
Limits
Why is there no GitHub Enterprise setting?
The client can be pointed at a different API root and the tests cover that, but there is no field for it in the UI yet. It is a small change if you need it.
Why no graphs?
GitHub traffic analytics live behind a separate account, and they have nothing to do with the core workflow. Deliberately left out.
Can it create API tokens for me?
No. Creating one needs two-factor re-authentication in a browser, so the app sends you to the token page instead of pretending it can do it.
Does it work offline?
Only with a local model. The GitHub side needs a network because nothing is cached — and that is on purpose, so you never see a stale view of a repository.
How many repositories can it see?
Up to 600, fetched page by page. With more than that you will see part of the list.
Troubleshooting
The AI buttons are greyed out
The AI provider is not configured. Go to Settings → AI and pick a provider. Ollama and LM Studio need no key — just the model, then Test connection.
The README comes back empty
Expected if the repository has no README yet. Use Generate README to create one, or Load existing to check the path is right.
I got a 403 rate limit error
Your hourly GitHub quota is spent. The app tells you roughly when it resets. It deliberately does not retry, because waiting a few seconds cannot help and would only freeze the interface.
The interface looks wrong
Set Windows display scaling to 100%, or nudge Settings → UI scale up or down.
Nothing happens after I press commit
The request is probably still running. A GitHub call can take a while. If it fails you get a message at the bottom of the page.
Contributing
The repository is open. If you find a problem or a missing feature:
- Open an issue
- Send a pull request if you like
Run the tests before you push:
python tests/run_all_tests.py
All nine suites should be green.
Licence
MIT. See >LICENSE in the repository.