Comparison · Same 38 public documents · Open benchmark
anymd vs MarkItDown
Both tools turn files into Markdown for LLMs, and both are free and MIT licensed. MarkItDown is Microsoft's Python library and CLI. anymd is one native Rust binary that is also an MCP server. On the 38 public documents of AgentDocBench, anymd scores 96.4 to MarkItDown's 76.8, converts all 38 in 11.5 s instead of 75.6 s, and returns 32% fewer output tokens.
Install anymd free See the full benchmark
npx -y @sylphx/anymd setup --client cursor,claude-code,codexAt a glance
| anymd | MarkItDown | |
|---|---|---|
| What it is | One native Rust binary: CLI and MCP server | Python library and CLI, with a separate MCP package |
| Install | npx -y @sylphx/anymd setup, Homebrew, Scoop, Cargo or Docker | pip install 'markitdown[all]', Python 3.10 to 3.14, virtual environment recommended |
| MCP tools | Built in: read, outline, search, inspect | markitdown-mcp: one tool, convert_to_markdown |
| AgentDocBench overall | 96.4 | 76.8 |
| Time, all 38 documents | 11.5 s | 75.6 s |
| Output tokens, all 38 documents | 368,659 | 538,388 |
| Scanned PDFs | Local OCR with tesseract, or a local document model after anymd setup ocr | No OCR by default; an OCR plugin that calls an LLM, or Azure Document Intelligence |
| Audio and video | Metadata, chapters, subtitles, and local speech-to-text after a one-time model download | Audio metadata and transcription, which its README says rely on web-hosted services; video through Azure Content Understanding |
| Python API | Prepared, not on PyPI yet | Yes |
| Inputs only MarkItDown reads | — | ZIP archives, Outlook messages, YouTube URLs |
| Plugins | None | Third-party plugins |
| Price | Free, MIT. Optional anymd Pro, one payment | Free, MIT |
Measured on the same 38 documents
AgentDocBench checks what an agent needs from a converter: the words verbatim, in reading order, with tables intact, fast, and in as few tokens as possible. Scores are 0 to 100; higher is better.
| AgentDocBench | anymd 8.2.0 | MarkItDown 0.1.8 |
|---|---|---|
| Overall | 96.4 | 76.8 |
| Documents converted | 38/38 | 38/38 |
| Sentences kept verbatim | 97.7 | 73.7 |
| Text F1 | 97.2 | 91.7 |
| Reading order | 98.8 | 85.5 |
| Table cells F1 | 92.2 | 57.2 |
| Time, all converted documents | 11.5 s | 75.6 s |
| Output tokens, all converted documents | 368,659 | 538,388 |
Document by document, anymd scores higher on 26, MarkItDown on 3, and 9 are tied.
By category
| Category | Documents | anymd | MarkItDown |
|---|---|---|---|
| Papers (math) | 3 | 96.7 | 74.2 |
| Two-column papers | 3 | 98.1 | 47.6 |
| Reports | 2 | 96.9 | 97.4 |
| Financial tables | 4 | 95.2 | 65.9 |
| Forms | 3 | 98.0 | 74.8 |
| Scanned (OCR) | 3 | 77.9 | 0.0 |
| CJK | 3 | 99.1 | 79.8 |
| Slides | 4 | 96.0 | 86.5 |
| Spreadsheets | 4 | 100.0 | 99.7 |
| Word | 3 | 99.9 | 99.8 |
| EPUB | 3 | 98.9 | 97.7 |
| HTML | 3 | 99.6 | 98.1 |
Where MarkItDown scores higher: Reports (97.4 vs 96.9); report-crs-dod-approps (95.8 vs 94.8); report-plz (99.1 vs 99.0); invoice-borderless (100.0 vs 94.8).
Leave out Scanned (OCR), where MarkItDown scores 0, and the overall is 98.0 for anymd and 83.8 for MarkItDown.
Why scanned PDFs score 0: with default options, MarkItDown returns no text for the three image-only PDFs. Its OCR plugin and Azure options were not measured.
How it was measured
Both tools ran on GitHub-hosted Linux x86_64 runners with 4 CPUs: MarkItDown 0.1.8 through its CLI with default options on 2026-09-25, and anymd 8.2.0 on 2026-10-01. Time is the wall time of a fresh process per file, start-up included, median of three runs. A long-running Python process would avoid part of MarkItDown's start-up time; that setup was not measured. Tokens are counted with o200k_base.
The benchmark lives in the anymd repository. The scoring does not know which tool produced an output, six of the 38 documents come from MarkItDown's own test files, and the corpus, ground truth, scoring and raw results are public in bench/. Anyone can rerun it or submit a tool.
Why agent builders switch to anymd
One binary, no Python environment
npx -y @sylphx/anymd setup adds anymd to Claude Code, Codex, Cursor, VS Code, Claude Desktop, Windsurf and Gemini CLI at once. Homebrew (brew install sylphxai/tap/anymd), Scoop (scoop install anymd from the SylphxAI/scoop-bucket bucket), Cargo (cargo install anymd) and Docker (ghcr.io/sylphxai/anymd) install the same binary. There is no virtual environment to keep in sync and no API key.
Layouts that usually break stay readable
anymd reads PDF glyph positions, so it keeps columns, tables and forms in order. The biggest gaps on AgentDocBench are two-column papers (98.1 vs 47.6), financial tables (95.2 vs 65.9), forms (98.0 vs 74.8) and CJK documents (99.1 vs 79.8). Table cells F1 is 92.2 vs 57.2.
Fewer tokens, more of the text
Across the 38 documents anymd returns 368,659 tokens to MarkItDown's 538,388, while keeping more sentences verbatim (97.7 vs 73.7). Its MCP tools add what an agent needs for long files: <!-- page N --> anchors to cite, outline to jump to a section, read with a token budget and a cursor, and search across a folder.
Your files stay on your machine
OCR and speech-to-text run locally. anymd goes online only when you pass a URL or explicitly ask it to download a model, and --allow-dir keeps the MCP server inside one folder.
When MarkItDown is the better fit
- You want an in-process Python library today. anymd's Python package is prepared but not yet on PyPI.
- You need ZIP archives, Outlook messages or YouTube URLs, or one of MarkItDown's third-party plugins.
- Your pipeline already uses Azure Document Intelligence or Azure Content Understanding.
Switch in one line
The CLI has the same shape:
markitdown report.pdf -o report.md # MarkItDown
anymd report.pdf -o report.md # anymd, after npm install -g @sylphx/anymdFor MCP, replace the markitdown-mcp entry in your client with npx -y @sylphx/anymd, or run npx -y @sylphx/anymd setup to add it to every client. With Docker:
docker run --rm -v "$PWD:/data" ghcr.io/sylphxai/anymd report.pdf > report.mdMore install options are in Getting started.
Check what your agent cites with anymd Pro
Converting a file is free. When an agent quotes a PDF or a video, anymd Pro takes you back to the source:
- Cite-check compares up to 100 quotes per request with the page and area each one cites, and reports whether it matches, doesn't match, or needs more evidence.
- Video timelines and frames show scenes, chapters and subtitle or transcript timings, and the frame behind a cited moment with its actual decoded timestamp.
Pay once, US$29 or the local price in 16 currencies, for a 100-year licence for one person. Pro works offline, with no subscription and no account. Prefer a hosted service? Use Sylphx AI parse to convert documents without installing a local runtime, and pay per page.
FAQ
Is anymd a drop-in replacement for MarkItDown? On the command line, mostly: anymd report.pdf -o report.md works like markitdown report.pdf -o report.md, and npx -y @sylphx/anymd setup adds anymd to your MCP clients. anymd's Python package is not on PyPI yet, so code that imports markitdown cannot switch today.
Is anymd faster than MarkItDown? On AgentDocBench, anymd converted all 38 documents in 11.5 seconds and MarkItDown in 75.6 seconds, each timed as a fresh process per file with start-up included.
Does anymd need Python? No. anymd is a native Rust binary. The npx install needs Node 18 or newer; Homebrew, Scoop and Docker install the same binary without Node.
Can MarkItDown read scanned PDFs? Not with its default options: on AgentDocBench its CLI returned no text for three image-only PDFs. It offers an OCR plugin that calls an LLM, and Azure Document Intelligence. anymd reads scans locally with tesseract, or with a local document model after anymd setup ocr, and scores 77.9 on the scanned category.
Is the benchmark biased toward anymd? It lives in the anymd repository, so you can check it: the scoring does not know which tool produced an output, the corpus and ground truth are public, six documents come from MarkItDown's own test files, and the categories where anymd loses are published. Anyone can rerun it or add a tool.
Is anymd free? Yes. anymd is MIT licensed, including the CLI, the MCP server, OCR and transcripts. anymd Pro is an optional one-time purchase, US$29 or the local price in 16 currencies, that adds PDF cite-check and video timelines and frames.
MarkItDown is a Microsoft project; anymd is not affiliated with or endorsed by Microsoft. Benchmark figures come from the AgentDocBench results committed in the anymd repository. MarkItDown details come from its README.