Receipt and document recognition
One photo is enough: merchant, amount, date, currency, category and line items arrive pre-filled in the form.
AI integration
We integrate Google Gemini, OpenAI and Anthropic Claude into apps and applications: read receipts, turn voice and photos into structured data – with clear rules for privacy.
Our approach
For us, AI is a tool with a clearly defined job, not an end in itself. The model delivers structured suggestions, and a person confirms them. In our apps users bring their own API key – the data goes straight from the device to the chosen provider, never through a server of ours.
Scope
One photo is enough: merchant, amount, date, currency, category and line items arrive pre-filled in the form.
Dictation or a photo of a note goes in, prioritised entries come out – for review before anything is saved.
One integration via the OpenAI-compatible chat completions API; Gemini, OpenAI and Claude differ only in base URL and model.
Clear statements about which data is transferred, and billing directly with the user’s provider.
Technology
The major language model providers, Apple’s speech recognition and secure key storage on the device.
Questions
Users store the API key of their own AI provider in the app. The key lives in the device keychain, requests go directly to the provider, and the provider bills the user for usage. We run no server for this and never see the data.
Only what the task needs – such as the receipt photo or the dictated text – and only when the feature is actively used. Without a configured provider nothing is transferred. We state this in the privacy policy and the App Store details as well.
Yes. There the integration runs on the server, for example in a Symfony application. We decide on provider, model and data flow together.