To publish documents as a static site, ask an AI agent to convert them to Markdown with YAML front matter and images saved as files, then run your generator. GroupDocs.Markdown.Mcp does the conversion locally, so Claude Desktop, Claude Code, Cursor or GitHub Copilot can prepare Word, Excel and e-book content for Hugo, Jekyll, MkDocs or Docusaurus without sending it to a service:

Convert these documents to Markdown with front matter, GitHub flavour, and images saved as files.

The step-by-step version with config and troubleshooting is in the documentation: How to publish documents to a static site as Markdown.

What does a static site generator need from a document?

Three things: Markdown for the body, YAML front matter for the metadata, and images as files at relative paths. Hugo, Jekyll, MkDocs and Docusaurus all read front matter. convert_to_markdown produces all three when you choose the options for it:

Option Value for a site What it does
frontMatter true Writes YAML with title, author, format and page count at the top of the file
images "file" Writes pictures next to the .md and references them by relative path
flavor "github" (default), or "commonmark" GitHub-flavoured Markdown is what most generators assume; CommonMark suits a stricter toolchain

Check the title and author at the top of each converted file; if they are empty or wrong, ask the agent to set them.

How do I convert Word to Hugo Markdown?

Hugo is the generator behind this blog, so it is the example here. The flow has four steps, and the agent runs all of them.

  1. Convert. Run the prompt above on the folder of documents.
  2. Arrange. The server writes files; the layout is the agent’s job.

    Put each converted file in a folder named after the document, with its images in an images subfolder next to it.

  3. Add the keys the theme wants. Hugo themes often expect date or weight. The tool does not write them, so ask:

    Add weight to the front matter of each file, numbered in the order I listed them.

  4. Build. Run Hugo yourself. At this point the files are ordinary Markdown with front matter, and the generator treats them like any page.

For Docusaurus and MkDocs the same output applies; only the folder layout and the extra keys differ, and you describe them in the second and third prompts.

Example session (abridged)

You:   Convert the three documents in /drafts to Markdown with front matter, GitHub
       flavour, and images as files. One folder per document.
Agent: convert_to_markdown (frontMatter: true, flavor: "github", images: "file") x 3
       Created guide-onboarding/, guide-security/, price-list/ with .md files and
       images subfolders. Front matter holds title, author, format and page count.
You:   Add weight to each file in the order I listed them.
Agent: Updated the front matter of the three files with weight 10, 20 and 30.

How do I check the result before I publish?

Two checks catch nearly every problem:

For each converted file, compare the page count of the source with the length of the Markdown.

A 60-page document that became 3 KB of Markdown is a scanned PDF or an evaluation-mode run.

List the image folder of each document and the images its Markdown references.

A .md that references pictures never written renders as broken images on the site. Spot-check the messiest source before a batch of two hundred.

Word and Excel sources convert with their structure intact: headings become headings and tables become Markdown tables. PDFs depend on how they were produced. A born-digital export converts cleanly; a scan does not convert in any useful sense, because this server has no OCR step.

Which front matter keys do I add myself?

frontMatter: true writes title, author, format and page count. Check those values in each file and ask the agent to correct any that are empty or wrong. Navigation, ordering, tags and dates are keys you add after conversion, as in step 3. To pull richer document properties into the front matter, pair this server with GroupDocs.Metadata.Mcp, which reads them, and have the agent add the keys; this post does not cover that.

Honest limits

Without a license the server runs in evaluation mode, and only the first 3 pages are processed, with an evaluation watermark on each. A site built from unlicensed output ships three-page stubs of long documents. Run get_license_status before you convert, and keep the “length against source” check in the loop.

FAQ

How do I convert Word to Hugo Markdown?

Ask the agent to convert the document with frontMatter: true, images: "file" and flavor: "github", then to arrange the files in your content folder. Add any theme-specific keys, such as weight, in a second prompt.

Can an AI agent publish documents to a static site?

It can prepare everything the site build needs: Markdown, front matter, image files and folder layout. You run the generator, so the build and the deployment stay under your control.

Does the conversion happen locally?

Yes. GroupDocs.Markdown.Mcp runs as a local process over stdio and reads the documents from the folder you configured. Documents that become public only after review stay on your machine until you publish them.

Go deeper