JustOnce deliberately runs no OCR and no document-reading model of its own. The model you already use — Claude, ChatGPT, Gemini — reads PDFs, scans and photos better than a bolt-on OCR pipeline, and it does so under your own account and terms. JustOnce's job is to remember what was read and to keep the original safely.
The workflow
Give the file to your assistant (drop it into the chat, or take a photo in your assistant's own app).
Ask: "Read this policy, save the key facts to JustOnce and keep the original attached."
The assistant transcribes what matters — vendor, dates, amounts, totals, reference numbers, names — and calls
store_documentwith that transcription plus the file. For a big original it gives you a one-time upload link instead (see the next article).
If only the file arrives
A PDF or image saved without a reading — for example through an upload link the assistant made without a transcription — becomes a title-only memory: the file is stored and downloadable, but there are no facts for recall to use. Fix it any time: ask your assistant to read the stored document and update the memory, or edit the memory in the app and type the facts in.
Exception: your own AI key
If you've added your own Anthropic, OpenAI or Google key under Settings → Email forwarding (paid plans), JustOnce uses it to read forwarded PDFs and images, and any PDF or image an assistant saves without a transcription. Reading with your own key doesn't use your pages.