mdstill
mdstill vs Unstructured

An Unstructured Alternative for When You Just Need the Markdown

If you need clean Markdown from a document right now -- a one-off file, a prototype, or moderate volume -- mdstill wins: it hands back token-aware GitHub-Flavored Markdown with zero setup, free and with no signup. Unstructured wins once ingestion is load-bearing: when you are an engineering team partitioning thousands of documents from many connected sources into typed elements and metadata for a vector DB at scale, and you want full control over that pipeline. Many teams use both -- mdstill for ad-hoc files and prototyping, Unstructured once the pipeline is the product.

upload_file

Drop file here or click to browse

…or paste a file from clipboard ⌘/Ctrl V

PDFDOCXDOCPPTXXLSXXLSHTMLHTMEPUBCSVJSONXMLRTFODTPAGESNUMBERSKEYNOTEZIP

Max file size: 20MB ·

mdstill vs Unstructured, side by side

AspectmdstillUnstructured
Primary purposeDocuments → ready-to-use Markdown for LLMs and RAGA document-ingestion toolkit/framework you build a pipeline with
SetupNone — drop a file in the browser, or one API callA Python environment and integration code, or an API key and wiring
OutputToken-aware GFM you can paste straight into a promptPartitioned elements / JSON you assemble into your own format
CostFree tier, no signup; Pro for higher limitsOpen-source you host, or a usage-priced hosted API
Connectors & sourcesSingle file upload (or one API request)Many source connectors built for ingestion at scale
Best forOne-off conversions, moderate volume, prototypingLarge-scale, multi-source document ingestion
PrivacyVolatile memory, deleted immediately, no retention or trainingSelf-hosted, or processed per their policy on the hosted API
Learning curveNone — the output is the productYou write integration code and tune the pipeline

Toolkit vs tool: what you are actually choosing

Unstructured is designed as infrastructure. It partitions a document into typed elements, attaches metadata, chunks for retrieval, and pulls from many connected sources -- which is exactly what you want when ingestion is a system you operate, not a step you run by hand. The cost of that power is that you assemble it: a Python pipeline or API calls, integration code, and the elements stitched into whatever your vector store expects. mdstill collapses that to one move -- file in, token-aware Markdown out -- which is the right trade when you need the result, not a framework.

When ready-to-use Markdown beats partitioned elements

A toolkit gives you structured elements precisely so you can build your own output. But for a one-off file, a prototype, or moderate volume, that assembly step is overhead you do not need. mdstill returns GitHub-Flavored Markdown directly: tables preserved as GFM pipe tables, H1–H6 heading hierarchy intact, and token count kept lean so the same document costs fewer tokens in a prompt. You paste it into ChatGPT or Claude, or drop the .md into a vector store, without writing a line of glue code first.

Privacy and the build-it-yourself question

Unstructured can be self-hosted, which gives an engineering team full control over where files live -- or you can use the hosted API, where documents are processed per their policy. mdstill takes a different stance: files are processed in volatile memory and deleted immediately after the conversion returns, with no logging of content, no retention, and no use for training. For sensitive one-off documents you do not want to stand up your own ingestion stack for, immediate deletion is the simpler safe default.

Use both — the honest workflow

These tools are not really rivals; they sit at different stages. Reach for mdstill while you are prototyping and converting ad-hoc files: it is instant, free, and needs nothing installed. Reach for Unstructured once the ingestion pipeline is load-bearing and pulls from many connected sources at scale. mdstill's developer-first API returns Markdown from a single multipart request, so even your scripted conversions stay a curl call away -- and the day you outgrow that for full pipeline control, Unstructured is where you graduate to. One note: mdstill targets documents and data formats and does not run OCR on scanned image-only PDFs.

When to use Unstructured instead

No single tool wins every job. Reach for Unstructured when:

Frequently asked questions

Related