Your documents.
A whole new workflow.✳

We have numbers. They come from a corpus we built ourselves, which makes them useless to you and dangerous to us. Run the free tier on your own documents and get one that means something.

Every migration rebuilds the layout from a picture of the page (visual reconstruction, Beta). From Word and RTF that picture is rendered from the file itself and cross-checked against its structure; PDFs and scans give us only the picture. Lower fidelity, closer review. Only the first 7 pages of a document are migrated — a longer one is truncated there, and we tell you it happened. Ruled scanned forms aren’t good enough yet.

Merged cells, images and links survive. Letterhead backgrounds drop; every list — flat or nested — exports as plain paragraphs with its markers written into the text, so Word won’t renumber it; a table’s per-cell borders and vertical alignment collapse to table-level styling; a ruled form keeps its text but not its drawn layout — labels and values print one under another; a block we have no .docx drawing path for becomes a labelled placeholder; and an Arabic or Hebrew paragraph is laid out left to right, so punctuation at its edges lands on the wrong side. PDF is the faithful deliverable, except for right-to-left and East Asian text — see Languages and scripts in the docs.

Sign where you sign today; call our API with your own data. If a connector directory or an on-prem package is a hard requirement, we’re honestly not your tool.

Word (.docx) and RTF — the higher-fidelity path, because the page is rendered from the file itself and cross-checked against its structure. PDF and image files are also accepted through the same visual reconstruction (Beta), with only the picture to go on, so review them more carefully. Only the first 7 pages of a document are migrated; a longer one is truncated there and the product says so.

The AI reports how sure it is on every field it labels. Fields it is unsure about flag themselves, so you know where to look first. It is usually right, not always: confidence can be wrong or miss a field, so check the complete template before you rely on it. (AI migration is in Beta.)

PDF and HTML generation is repeatable with the same template, data, assets and renderer version. No AI runs at generation time. DOCX output is editable and can differ in its bytes between renders. Test templates with representative data before automating them.

Yes — a keyed /v1 API generates single documents and batches. See the API docs.

Check Settings → Usage any time to see how close you are. Migration, recognition and generation all pause at the credit cap until the next monthly cycle or an upgrade — or you can buy a one-time credit pack to keep going immediately. You're never billed for usage you didn't buy.

We minimise diagnostic data and mask known sensitive patterns; unexpected personal data can still enter an error report. AI migration sends page images and reconstructed text to Anthropic (processing in the US; Anthropic Ireland, Limited contracts for EEA customers). Use ordinary business documents only, without excluded sensitive data. See /privacy and /subprocessors for recipients, retention and safeguards.