A cookbook that works offline
Saved recipes are cached on the device: search, scale servings up or down, and work through cook-mode checklists with no signal in the kitchen.
Independent software studio
Two founders, one project at a time: a cooking app that turns a spoken recipe into a structured card you can edit, save and cook from.
Talk through a recipe the way you would tell it to someone in your kitchen. Kūra transcribes what you said and pulls out the ingredients, steps, timings, tags and allergens as a card. The model drafts it; you correct it inline before it is saved. Typing gives you the same card, minus the microphone.
Learn and love to cook — without losing the recipe your grandmother only ever said out loud.
The K is the navigation
What it does
Saved recipes are cached on the device: search, scale servings up or down, and work through cook-mode checklists with no signal in the kitchen.
Add items by typing them, or photograph the shelf and let vision identify what is there.
Dishes ranked by how much of the recipe your pantry already covers, linked back to the matching recipes you have saved.
Every edit you make to a generated card is diffed and stored, so the model gets better at the kind of cooking you actually do.
How it works
No API keys ship in the app bundle. The client holds only the public Supabase anon key; the Deepgram and Gemini keys live as Edge Function secrets, and every function verifies the caller’s token before it does any work.
Built with
Studio
exebit is an equal partnership between its two founders. Both of us are engineers and both of us write the code that ships; the titles describe who signs what, not a split in the work.
We keep one project in front of us at a time so it actually gets finished. To talk about work, kūra, or anything adjacent, the address below reaches both of us.
Contact
hello@exebit.lolEnquiries, support, or early access to kūra. We answer within two business days.