KoeNote

You open the coffee app. You fill in the form. You open the journal app. You fill in the form. You open the spreadsheet. You fill in the form. People said LLMs would change how we interact with software. The form is still there. KoeNote changes that. Tap record. Speak naturally. KoeNote transcribes your voice, sends it through an AI model using a template you define, and saves the structured result directly to your GitHub repository — in seconds. Your coffee tasting note becomes a formatted json file. Your workout log becomes a structured entry. Your journal becomes a timestamped record. No forms. No friction. Just speak and move on. What makes it different KoeNote is not a notes app. It is a capture layer. Your GitHub repository is the destination — what you build on top of it is entirely up to you. Connect it to Obsidian, trigger a GitHub Action, build a personal dashboard, pipe it into a database. KoeNote gets the data in. You decide what happens next. Templates you own Templates are simple instructions you write once: "Format this as a coffee brewing log with fields for dose, ratio, and tasting notes." KoeNote applies your template to every recording. Different activity, different template — the output is always exactly what you defined. Don't want to write one from scratch? Just speak what you want to capture — KoeNote generates the template for you instantly. Already have a record and want to add more? Speak your update and KoeNote merges it into the existing entry. Your data stays complete, not scattered across separate files. Your data, your keys KoeNote has no backend. There is no KoeNote account. Your API keys for Claude and OpenAI are stored exclusively in your device's Keychain — never transmitted to or stored by KoeNote. Your GitHub personal access token lives in the same place. Nothing leaves your device except the API calls you explicitly authorize. Set spending limits directly in your LLM provider's dashboard. KoeNote never touches billing — your keys, your budget, your control. What you need to get started - A GitHub account and a personal access token (with `contents: write` permission on your target repo) - An API key from Anthropic (Claude) or OpenAI - About five minutes to set up your first template Supported LLM providers - Anthropic Claude (`claude-haiku-4-5` default — fast and cost-efficient for structured extraction) - OpenAI (`gpt-5.4-nano` default — fast and cost-efficient for structured extraction) You can change the model name in Settings. Google Gemini support is planned for a future release.