Is there a hosted MarkItDown API?

flatmark's direct call, POST /v1/convert, runs MarkItDown on files up to 8 MB. It adds an API key, rate limits, credits (1 per keyed call) and an MCP server.

Why not pip install markitdown?

If you can run Python next to your files, do. MarkItDown is MIT-licensed, and pip install 'markitdown[all]' brings every optional converter (source: github.com/microsoft/markitdown, as of Sep 2026). The API helps where you cannot install packages.

Does flatmark only use MarkItDown?

No. The direct call runs MarkItDown. The queue runs Docling, which adds OCR, PDF headings and a JSON structure file.

flatmark vs MarkItDown

MarkItDown is Microsoft's open-source file-to-Markdown converter and the engine behind flatmark's direct call. A local library compared with a hosted API.

Last updated Sep 24, 2026

MarkItDown	Open-source (MIT) Python tool from Microsoft. It converts PDF, Word, PowerPoint, Excel, HTML and other files to Markdown (as of Sep 2026)
Install	pip install 'markitdown[all]' for all optional converters (as of Sep 2026)
PDF OCR	The built-in PDF converter does no OCR. The markitdown-ocr plugin adds it through an LLM vision model (as of Sep 2026)
Relationship	flatmark's direct call (POST /v1/convert) runs MarkItDown on files up to 8 MB

MarkItDown is Microsoft's open-source Python tool for converting files to Markdown, under the MIT license. Install it with pip install 'markitdown[all]'; a bare pip install markitdown leaves out the optional dependencies (GitHub, as of Sep 2026). Its README says it "may not be the best option for high-fidelity document conversions for human consumption" (same source).

flatmark's direct call runs MarkItDown. POST /v1/convert takes files up to 8 MB and returns JSON with the Markdown. Anonymous calls are allowed; keyed calls cost 1 credit.

When MarkItDown fits

You work in Python and the files sit next to your code. A local dependency then costs nothing per call and needs no network. For scripts and notebooks it is the shorter path.

When flatmark fits

You are not in Python, or conversion runs where you cannot install packages: a serverless function, an agent tool call, a browser client. The conversion is then one HTTP call or MCP tool call.

PDFs from the direct call come back as text without headings and without OCR, as with MarkItDown itself. For scans and PDF headings, use the queue (POST /v1/convert/jobs, Docling, up to 25 MB and 200 pages). It is a separate endpoint with its own polling flow.

Get an API key with a magic link. There is no password to manage.

FAQ

Is there a hosted MarkItDown API?: flatmark's direct call, POST /v1/convert, runs MarkItDown on files up to 8 MB. It adds an API key, rate limits, credits (1 per keyed call) and an MCP server.
Why not pip install markitdown?: If you can run Python next to your files, do. MarkItDown is MIT-licensed, and pip install 'markitdown[all]' brings every optional converter (source: github.com/microsoft/markitdown, as of Sep 2026). The API helps where you cannot install packages.
Does flatmark only use MarkItDown?: No. The direct call runs MarkItDown. The queue runs Docling, which adds OCR, PDF headings and a JSON structure file.