Handwriting OCR

Create a single-use URL to upload a file from disk, for clients that can run shell commands (for example Claude Code). Use this instead of base64 in submit-document whenever the file is on disk. Send the file with the returned curl command; the upload creates the document, starts processing and returns a document_id, exactly as submit-document does. The URL expires after 10 minutes and works once. Each processed page uses one page credit, so ask before uploading many files.

create-upload-url

Export a processed document to a formatted file and get a temporary download URL. The document must already be processed. Use get-document first if unsure of its status.

export-document

Get the processing status of a document by its id, and its results once status is "processed". Poll this after submit-document.

get-document

List your recent documents with their processing status and id, newest first.

list-documents

List your extractors (id, name, description, status) for use with the "extractor" action of submit-document. Only extractors with status "ready" can be submitted; an "incomplete" one needs to be tested in the extractor editor first.

list-extractors

Submit a PDF or image for processing. Use file for an attached ChatGPT file, or filename and base64 content in other clients. Base64 is for small files only: when you can run shell commands and the file is on disk, use create-upload-url instead. Returns a document_id; poll get-document for results. Each processed page uses one page credit. Each call creates a new document, so ask before retrying a submission.

submit-document