CambioML

AnyParser is an API and a Python SDK for reading documents that plain text extraction tends to mangle. Send it a PDF, a Word file, a slide deck, or a photo of a printed page, and it gives back the content as Markdown with headings and reading order intact. Ask for tables and you get HTML. Ask for fields and you get JSON.

CambioML builds the parser on vision language models instead of the usual OCR plus template pipeline. That choice explains most of what the product does well. A scanned invoice and a born-digital report go through the same call, and you don't write a layout rule for each new document type. For a team feeding messy paperwork into a retrieval system, that removes a lot of setup work.

The limits show up in three places. Output is a model reading of the page, so dense tables and faint scans still need a spot check before the numbers land in a database. The standalone anyparser.com site currently shows a coming soon page, and CambioML's main site now leads with insurance document automation, so the documentation site is the reliable entry point. Pricing above the free tier runs through sales rather than a checkout page, which makes cost planning slower than with self-serve tools.

Getting Started

1. Create an API key on the CambioML account page. Every account starts with 100 free pages, and the free key handles up to 10 pages per call.
2. Install the SDK with pip install any-parser and keep the key in an environment variable such as CAMBIO_API_KEY. The REST API is there if you would rather skip Python.
3. Parse a short file synchronously with ap.parse(file_path=...). The sync endpoint gives up after 30 seconds, which suits one or two page documents.
4. Switch to async_parse plus async_fetch for longer files. The first call returns a file ID, and the second pulls the Markdown when processing finishes.
5. Choose the output you need: full Markdown, HTML tables, key-value JSON, or PII. Check pageRemaining through the /query endpoint to keep an eye on your balance.