What comes across
Every .html file in the .zip of a space's HTML export is read. By default the space becomes a folder of pages: a Markdown document per page, a subfolder per section, the section's own page first inside it. One document joins the pages into one file instead.
| In the export | In the Markdown |
|---|---|
The title, written Space : Page |
# Page, the space name taken off |
| The breadcrumbs | The folder the page is filed in |
| Headings, lists, bold, italic | The same, in Markdown |
| A table | A Markdown table, its first row as the header |
| The contents macro, a link to a heading | A link to the heading's GitHub-style slug |
| A link to another page | A link to that page, in a folder of pages |
A picture under attachments/ |
Carried inside the page, a heavy one first drawn down to at most 1,600 pixels on its long side |
| A link to a web address | Kept as a link |
What is left behind
- Macros as macros. None has a rule of its own: each arrives as the HTML the export made of it; an info panel is just its paragraphs.
- Merged cells. The text lands in the first cell it covered; below a cell merged downwards the row shifts left. A cell holding several paragraphs or a list is flattened onto one line.
- Attachments that are not pictures. A PDF is not carried, so in a folder of pages its link becomes its name, as do links to Confluence's own screens.
- Pictures past the allowance. In a folder of pages each may carry nearly 4 MB of them, none over 1 MB encoded; one that does not fit becomes its description or file name in italics. One document shares 2 MB; a picture past it keeps its
attachments/path.
An example
A space export in which "Deploy checklist", under "Runbooks", holds a contents macro, a heading, a table, a link to the "Rollback" page, a linked runbook.pdf and a screenshot. As a folder of pages, downloaded as a .zip, it becomes Engineering/Runbooks/Deploy checklist.md (picture data shortened, page frame left out):
# Deploy checklist
- [Before you deploy](#before-you-deploy)
## Before you deploy
| Step | Owner |
| --- | --- |
| Tag the release | Ops |
| Announce the freeze | Release lead |
If it fails, follow [Rollback](Rollback.md). The full runbook is runbook.pdf.
The contents entry pointed at Confluence's id Deploychecklist-Beforeyoudeploy and now jumps by the slug. As one document, the page would be a section under a # Contents list of titles.
Limits
The .zip may be up to 250 MB; only its pages and pictures are unpacked, and one with no .html files is refused. It is converted in your browser, and nothing is uploaded until you sign in and save. A saved page may be up to 4 MB.
Pages are ordered by file name, index.html first, not by Confluence's page tree. Nothing picks out the main content: beyond the breadcrumbs, the page frame comes across, such as a heading repeating the space name, an author line or a "Document generated by Confluence" line. Code becomes a fenced block only where the HTML has <code> inside <pre>; a bare <pre> arrives as plain lines. Ticks are read only from HTML checkboxes, not from Confluence's task lists.
More on macros in converting a Confluence page to Markdown, and on the .zip in how to open an export .zip.
Questions
Is this a Confluence app or plugin?
No. Nothing is installed in Confluence or connects to it: you export the space as HTML and drop the .zip here. The free options compared covers the Marketplace apps.
Does it work with exports from Confluence Cloud and Data Center?
Neither has a separate path: from any HTML export the converter reads each page's <title>, breadcrumbs, heading ids and file name, and it takes only that export's .zip.
Can I convert a single Confluence page to Markdown?
Yes, if it arrives as HTML in a .zip: one page converts the same way as six hundred.
Can I get HTML instead of Markdown?
Choose One document: its download menu offers HTML, Word and plain text too. A folder of pages downloads as Markdown only.
Do links between pages keep working?
In a folder of pages, yes: each points at the saved document in your account, or at its file in the downloaded .zip. In one document they keep the export's file names, which lead nowhere outside it.