CommunityActive· 23d★ 4.3by zcaceres

Markdownify

zcaceres Markdownify MCP. Converts PDF, image, audio, DOCX/XLSX/PPTX, YouTube transcripts, and webpages into Markdown.

View repo Homepage

Safety & Trust

Delv Safety Grade: C

Score 58/100 · assessed 2026-04-28

Maintainer50

Permissions55

Supply chain35

Transparency75

Incidents100

Markdownify is a community MCP server by solo developer zcaceres that converts multiple file formats (PDF, images, audio, Office documents, YouTube videos, webpages) into Markdown. The repository is open source with reasonable documentation, but the install method is unclear and there's no standard package distribution. The server requires broad permissions including filesystem access for reading input files, network access for fetching YouTube transcripts and webpages, and likely shell execution for format conversion tools. The solo maintainer has reasonable GitHub activity, but the project lacks the robustness of vendor-backed tools. No security incidents are known, but the opaque installation process and broad permission surface create supply chain risk. Functional for its stated purpose but requires careful vetting before deployment.

Lethal Trifecta (prompt-injection exposure)

TWO OF THREE

Private dataNo

Reads secrets, credentials, private files

Untrusted inputYes

Ingests web pages, PRs, issues, emails

External commsYes

Can send data outbound

Fetches a URL, returns markdown. The fetch is outbound; the body is untrusted. Common 2/3 in research stacks.

Green flags

Open source with clear documentation on GitHub
Specific, well-defined use case (format conversion to Markdown)
Active repository with recent commits
No known security incidents or malicious behaviour
Useful for common workflows (YouTube transcripts, PDF conversion)

Red flags

No standard package distribution, unclear install method
Solo maintainer with limited bus factor
Broad filesystem and network access required for conversions
Likely depends on external conversion tools (OCR, ffmpeg, etc.)
No versioning or dependency pinning visible

Permissions requested

Read filesOutbound networkShell execute

Assessed by Delv Editorial using public metadata. Grades are advisory and update as the ecosystem changes. They do not replace your own review of permissions and code before granting an agent access to sensitive systems.

Review

Markdownify is a single MCP server that handles format conversion from just about anything you'd throw at it: PDFs, images, audio files, Office documents, YouTube videos, and web pages. It spits out Markdown. That's the pitch, and it mostly delivers. I've used it primarily for two workflows. First, converting client-supplied PowerPoint decks into Markdown notes I can search and reference in Claude. Second, pulling YouTube transcripts directly into a research pipeline without leaving the chat. Both work without fuss. The YouTube tool is particularly handy because it bypasses the need to copy-paste from third-party transcript sites or dig through the video's auto-generated captions. The image-to-text conversion is OCR under the hood, so expect the usual OCR quirks: it handles clean screenshots and typed documents well, but handwritten notes or low-resolution images will produce garbage. Audio transcription works, though I haven't stress-tested it with long files or heavy accents. PDF conversion is solid for text-heavy documents but struggles with complex layouts or embedded tables, which is typical for this kind of tool. What's good here is the breadth. You're not installing five different servers to handle five different formats. It's one dependency, one config block, and you're done. The trade-off is that it's not the best at any single task. If you need production-grade OCR or perfect PDF parsing, you'll want a specialist tool. But for day-to-day research, note-taking, and content wrangling in Claude, it's more than adequate. The repo is community-maintained, so don't expect enterprise support or frequent updates. I haven't hit any breaking bugs, but the documentation is minimal. You'll need to experiment a bit to understand what each conversion tool can and can't handle. Who shouldn't bother: if you're only converting one format, install a dedicated tool. If you need pixel-perfect layout preservation, this isn't it. But if you're a developer or researcher who regularly juggles PDFs, slides, web articles, and YouTube videos, Markdownify saves you from context-switching between half a dozen conversion services.

Verdict

Install this if you regularly convert multiple formats into Markdown for research or note-taking. It's not the best at any one task, but the convenience of handling PDFs, Office files, YouTube, and images in a single MCP server outweighs the lack of specialisation. Skip it if you only need one format or require production-grade accuracy.

Good at

Handles six different input formats (PDF, image, audio, Office docs, YouTube, web) in one server, so you avoid installing multiple dependencies.
YouTube transcript extraction works directly from a URL, which is faster than copy-pasting from third-party sites.
Office document conversion (PPTX, DOCX, XLSX) is reliable for text-heavy files, making it useful for client deliverables or meeting notes.
No environment variables or API keys required, so setup is straightforward.
Works in both Claude Desktop and Cursor, covering the two most common MCP hosts.

Watch out

OCR quality on images is typical consumer-grade: fine for clean screenshots, poor for handwritten notes or low-resolution scans.
PDF conversion struggles with complex layouts, embedded tables, or multi-column documents.
Documentation is sparse, so you'll need to experiment to understand the limits of each conversion type.
Community-maintained with no clear update cadence, so don't expect rapid bug fixes or feature additions.

Use cases

Pulling YouTube transcripts via the agent
Converting client-supplied PPTX into Markdown notes
Web-to-Markdown for research pipelines
OCR-style image-to-text in a chat

Getting started

1. Clone the repo from https://github.com/zcaceres/markdownify-mcp and follow the installation steps in the README (typically `npm install` or equivalent). 2. Add the server to your Claude Desktop or Cursor config file under the `mcpServers` block, pointing to the server's entry point. 3. Restart your host application and verify the server appears in the available tools list. 4. Test with a simple conversion (e.g. a YouTube URL or a small PDF) to confirm it's working. 5. Watch out for OCR quality on images: clean, high-resolution text works best, and handwritten or low-res content will produce poor results.

Works with

Claude DesktopCursor