MarkItDown

Convert PDF, Word & More to Markdown

Convert PDFs, Office files, tables, images, and URLs into clean Markdown for AI workflows—locally, on the server, or with OCR when needed.

-[x]Free to start-[x]Secure processing-[x]Easy Markdown export

// try it now

Convert Your Files with MarkItDown

Upload a file or fetch a public URL, then preview and download .md output in seconds.

// overview

What Is MarkItDown?

MarkItDown is an online converter inspired by Microsoft's open-source utility. It handles PDF, DOCX, XLS/XLSX, PPTX, HTML/HTM, CSV/TSV, JSON, XML, TXT/MD, EPUB, IPYNB, public URLs, and PNG/JPG/WebP OCR, then prepares the content for LLM prompts, RAG, search indexes, docs, and Git workflows.

Built for AI Workflows

Send clean output into RAG pipelines, LLM prompts, vector databases, and knowledge bases.

Preserves Document Structure

Headings, tables, lists, links, and inline formatting stay readable after conversion.

Open Source and Extensible

The original project is MIT-licensed; this site self-hosts the Microsoft MarkItDown service.

// examples

See What MarkItDown Actually Produces

Inspect real inputs and raw output before choosing a mode: local, server, OCR, math, or diagrams.

Open all examples

PPTX deckworkshop-pitch.cloud.md

### Notes:Mention the waitlist numbers from the attendance sheet if asked.<!-- Slide number: 3 -->Program economics| Line | Monthly | Covered by || --- | --- | --- || Venue hire | $420 | Council grant || Materials | $300 avg | Ticket sales || Instructor fees | $650 avg | This proposal |

PPTX deck

Slides, speaker notes, and a real table

Compare local deck extraction with server MarkItDown on comments and a three-column table.

Browser localCloud MarkItDown

Open sample

Receipt OCRworkshop-receipt.ai-ocr.md

HARBOR TOOL LIBRARYWorkshop receipt - 18 August 2026Order: W-1842    Member: A. RiveraItem                              Qty          Unit          TotalHardwood sheet                    2            $18.00        $36.00Safety gloves                     3            $7.50         $22.50Router bit rental                 1            $14.00        $14.00

Receipt OCR

Readable image text without pretending it is accounting data

See when Local OCR is enough and when AI OCR keeps fields and totals cleaner.

Local OCRAI OCR

Open sample

Math OCRmoving-averages-scan.ai-ocr.md

1. DefinitionGiven a series x₁, ..., xₙ and weights wᵢ, the weighted moving average over a window of size k ismₜ = (∑ᵢ₌₀ᵏ⁻¹ wᵢ xₜ₋ᵢ) / (∑ᵢ₌₀ᵏ⁻¹ wᵢ)2. Variance Under IndependenceIf the xᵢ are independent with common variance σ², the variance of the estimator isVar(mₜ) = σ² · (∑ wᵢ²) / (∑ wᵢ)²

Math OCR

Equations stay reviewable when ordinary OCR breaks down

A scanned handout shows how OCR handles variables and symbols.

Scanned PDFAI OCR

Open sample

Pro renderingab-test-note.md

## Confidence interval for a conversion rateWith $n$ visitors and $k$ conversions the point estimate is $\hat{p} = k/n$, and the normal-approximation interval at level $1-\alpha$ is$$\hat{p} \pm z_{1-\alpha/2} \sqrt{\frac{\hat{p}(1-\hat{p})}{n}}$$## Required sample sizeTo detect an absolute lift of $\delta$ with power $1-\beta$, each arm needs roughly

Pro rendering

LaTeX and Mermaid remain portable source

Preview rendered math and diagrams without hiding the source code.

KaTeXMermaid

Open sample

// markitdown paths

Choose a MarkItDown Path

Pick the source type and open the focused MarkItDown tool.

// documents

Documents

PDF to Markdown Word to Markdown PPT to Markdown Excel to Markdown

// data

Data & Tables

CSV to Markdown Table to Markdown JSON to Markdown TXT to Markdown

// web + images

Web & Images

URL to Markdown HTML to Markdown Image to Markdown

// write + export

Output Tools

Markdown Viewer Markdown to PDF Markdown to Word Markdown Cheatsheet

// how it works

How MarkItDown Chooses the Right Conversion Path

Start with the file in front of you: local for supported files, server for richer parsing, OCR for scans, then inspect, edit, and export.

// route

Local or server is a visible choice

source.pdf

choose path

local browser

Supported files stay on device.

server markitdown

Richer parser for complex formats.

LocalServerURL

The processing path is explicit before the file runs.

// capabilities

Why Choose MarkItDown?

MarkItDown is not just a file drop zone. It gives clear paths from messy files to output you can inspect, edit, save, and reuse.

One Workspace for Real File Mixes

PDF, Word, PowerPoint, Excel, HTML, CSV/TSV, JSON/XML, EPUB, notebooks, public URLs, pasted tables, and images can all enter one workflow.

Local OCR and High-Accuracy AI OCR

Scanned PDFs and PNG/JPG/WebP images can start with Local OCR. Dense tables, totals, formulas, or small text can switch to AI OCR.

Free to Start, Useful Before Sign-In

Start without installing Python or creating an account. Sign in only for saved results, batch work, larger files, or more OCR capacity.

Output You Can Inspect

The raw result is visible, not hidden behind a polished preview. Check tables, headings, links, code, LaTeX, and Mermaid before handoff.

Browser-Local or Cloud MarkItDown

Use Local when supported files should stay on-device. Switch to Cloud MarkItDown for richer parsing, server-only formats, or comparison.

Edit, Export, and Keep the Boundary Clear

Open the result in the editor, copy it, download .md, or save it. Local, server, and OCR paths stay clearly labeled.

Frequently Asked Questions About MarkItDown

Key details about supported formats, OCR, privacy, limits, and project use.

What is MarkItDown?

MarkItDown began as Microsoft's open-source utility for extracting structured text from many file formats. This online version focuses on PDF, DOCX, PPTX, XLS/XLSX, HTML/HTM, TXT/MD, CSV/TSV, JSON/XML, EPUB, IPYNB, public URLs, and PNG/JPG/WebP OCR.

What file formats does MarkItDown support?

The online converter supports PDF, DOCX, PPTX, XLS/XLSX, HTML/HTM, TXT/MD, CSV/TSV, JSON, XML, EPUB, IPYNB, public URLs, pasted tables, and PNG/JPG/WebP images through OCR. EPUB and IPYNB use the server path; image formats use OCR.

Is MarkItDown free to use?

Yes. You can start without an account. The free tier handles up to 3 files per batch and files up to 20MB. Pro raises limits to 100MB files, 50-file batches, more OCR, and unlimited saved results.

How does MarkItDown handle PDF files?

MarkItDown extracts selectable PDF text and preserves structure where possible. Scanned PDFs can use AI OCR, including table recovery and math converted to LaTeX. Free accounts include 10 OCR pages per month.

Can MarkItDown preserve tables and formatting?

Yes, when the source exposes that structure clearly. Tables, headings, lists, links, and inline formatting are retained where the parser can read them.

Is MarkItDown safe for sensitive documents?

Uploaded files pass through Cloudflare to our self-hosted Microsoft MarkItDown service, are processed as temporary files, and are deleted after conversion. We do not store source files or log file contents; saved results are stored only when you choose to save them.

Can I use MarkItDown in my own projects?

Yes. Microsoft's MarkItDown is available as an open-source Python package for applications, scripts, data pipelines, CI jobs, and batch processing. See the GitHub repository for installation and API examples.

More resources: Microsoft MarkItDown on GitHub, MarkItDown with Claude, MarkItDown MCP, pricing, and the blog.

// get started

Start Converting with MarkItDown Today

Turn source files into structured output for AI, research, docs, and publishing with clear processing boundaries.

Start Converting View on GitHub