Comparison · Same 38 public documents · Open benchmark
anymd vs Unstructured
Both tools prepare documents for LLM and RAG pipelines. Unstructured is an Apache 2.0 Python library that splits files into typed elements, with connectors for batch ingestion. anymd is one native Rust binary, MIT licensed, that returns Markdown and is also an MCP server. On the 38 public documents of AgentDocBench, anymd scores 96.4 to Unstructured's 81.2, converts all 38 in 11.5 s instead of 346.5 s, and returns 11% fewer output tokens. Tables make the biggest difference: table cells F1 is 92.2 against 38.4.
Install anymd free See the full benchmark
npx -y @sylphx/anymd setup --client cursor,claude-code,codexAt a glance
| anymd | Unstructured | |
|---|---|---|
| What it is | One native Rust binary: CLI and MCP server | Python library: partition() splits files into typed elements; batch connectors in unstructured-ingest |
| Install | npx -y @sylphx/anymd setup, Homebrew, Scoop, Cargo or Docker | pip install "unstructured[all-docs]" on Python 3.11 to 3.13, plus poppler and tesseract for PDFs and images and LibreOffice for Office files |
| MCP tools | Built in: read, outline, search, inspect | Not part of the library |
| AgentDocBench overall | 96.4 | 81.2 |
| Table cells F1 | 92.2 | 38.4 |
| Time, all 38 documents, 4 CPUs | 11.5 s | 346.5 s |
| Output tokens, all 38 documents | 368,659 | 414,333 |
| Output | Markdown with <!-- page N --> anchors; JSON with geometry through inspect | Elements (such as Title, NarrativeText, Table) with metadata, and Markdown through elements_to_md |
| For RAG pipelines | Token-budgeted read with a cursor, outline and search | Chunking (for example by title), and source and destination connectors |
| Usage analytics | None | By default, a ping when the library loads and on each partition call, per its current README; set DO_NOT_TRACK to opt out |
| Licence | MIT | Apache 2.0 |
| Price | Free, MIT. Optional anymd Pro, one payment | Free library; the hosted Transform API is a separate service |
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 | Unstructured 0.27.8 |
|---|---|---|
| Overall | 96.4 | 81.2 |
| Documents converted | 38/38 | 38/38 |
| Sentences kept verbatim | 97.7 | 92.0 |
| Text F1 | 97.2 | 97.1 |
| Reading order | 98.8 | 93.9 |
| Table cells F1 | 92.2 | 38.4 |
| Time, all converted documents | 11.5 s | 346.5 s |
| Output tokens, all converted documents | 368,659 | 414,333 |
Document by document, anymd scores higher on 26, Unstructured on 4, and 8 are tied.
By category
| Category | Documents | anymd | Unstructured |
|---|---|---|---|
| Papers (math) | 3 | 96.7 | 80.3 |
| Two-column papers | 3 | 98.1 | 60.8 |
| Reports | 2 | 96.9 | 87.4 |
| Financial tables | 4 | 95.2 | 71.9 |
| Forms | 3 | 98.0 | 74.6 |
| Scanned (OCR) | 3 | 77.9 | 60.5 |
| CJK | 3 | 99.1 | 71.8 |
| Slides | 4 | 96.0 | 84.9 |
| Spreadsheets | 4 | 100.0 | 85.7 |
| Word | 3 | 99.9 | 99.8 |
| EPUB | 3 | 98.9 | 99.1 |
| HTML | 3 | 99.6 | 98.1 |
Where Unstructured scores higher: EPUB (99.1 vs 98.9); paper-rankone (95.9 vs 94.8); report-plz (99.9 vs 99.0); slides-markitdown-pptx (100.0 vs 98.5); epub-markitdown (97.2 vs 96.8).
The words are there; the structure is not. Text F1 is almost the same (97.2 vs 97.1), but with its default strategy Unstructured keeps few table cells in place (38.4) and scores 60.8 on two-column papers, so an agent gets the right words in the wrong shape.
How it was measured
Both tools ran on GitHub-hosted Linux x86_64 runners with 4 CPUs: Unstructured 0.27.8 on 2026-09-25, through partition() with its default auto strategy, which does not use the hi_res layout model, and then elements_to_md(); and anymd 8.2.0 on 2026-10-01. Time is the wall time of a fresh process per file, start-up included: one timed run for Unstructured, and the median of three runs for anymd. The hi_res strategy, the hosted Transform API and Unstructured 0.27.37, released after the run, were 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, and the corpus, ground truth, scoring and raw results are public in bench/. Anyone can rerun it or submit a tool.
Why agent builders pick anymd
One binary instead of a Python stack
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, and no poppler or LibreOffice to install: Office files, EPUB and HTML are parsed natively.
Tables and columns stay intact
anymd rebuilds tables from ruling lines and aligned whitespace and reads two-column pages in order. The largest category gaps are two-column papers (98.1 vs 60.8), CJK documents (99.1 vs 71.8), financial tables (95.2 vs 71.9) and forms (98.0 vs 74.6).
30× faster, in fewer tokens
On the same 4-CPU runner, anymd converted the 38 documents in 11.5 s and Unstructured in 346.5 s, and anymd's output was 368,659 tokens to Unstructured's 414,333. Pages come back with <!-- page N --> anchors, so an agent can cite where an answer came from.
Nothing leaves your machine
anymd sends no usage analytics. It 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 Unstructured is the better fit
- You want typed elements with metadata and built-in chunking inside a Python RAG pipeline.
- You ingest from many sources and destinations through its connectors.
- You plan to use its
hi_reslayout model or the hosted Transform API, which this benchmark did not measure. - Your corpus is mostly EPUB, where Unstructured scored slightly higher (99.1 vs 98.9).
Switch in one line
# Unstructured
from unstructured.partition.auto import partition
from unstructured.staging.base import elements_to_md
markdown = elements_to_md(partition(filename="report.pdf"))anymd report.pdf -o report.md # anymd, after npm install -g @sylphx/anymdFor MCP, add npx -y @sylphx/anymd to your client, 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. It uses the text and its position on the page, not an AI opinion.
- Video timelines and frames show scenes, chapters and subtitle or transcript timings, and the frame behind a cited moment with its actual decoded timestamp and a SHA-256 hash of the image.
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 Unstructured? For Markdown output from the command line or an MCP client, yes: anymd report.pdf -o report.md, or npx -y @sylphx/anymd setup for your MCP clients. Code that calls partition() and works with elements cannot switch today, because anymd's Python package is not on PyPI yet and anymd returns Markdown, not elements.
Is anymd faster than Unstructured? On AgentDocBench, anymd converted all 38 documents in 11.5 seconds and Unstructured in 346.5 seconds on the same 4-CPU runner, each timed as a fresh process per file with start-up included.
Why is Unstructured's table score low? With its default auto strategy, Unstructured kept few table cells in their rows and columns: table cells F1 is 38.4 against anymd's 92.2. Its hi_res strategy, which uses a layout model, was not measured.
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.
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, 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.
Unstructured is an open-source project of Unstructured (unstructured.io); anymd is not affiliated with or endorsed by it. Benchmark figures come from the AgentDocBench results committed in the anymd repository. Unstructured details come from its README and documentation.