Export content
Export a publication target's content as files you can take elsewhere: Markdown, DITA, HDITA, Word, XLIFF, OKF, a RAG bundle, or PDF. Everything runs from one dialog.
Markdown and PDF export are available on every plan. DITA, HDITA, Word, XLIFF, OKF, and the RAG bundle require the Business plan. Gated rows show Upgrade in place of the download button.
Open the Export dialog
Go to Publishing in the sidebar and open the map you want to export, or open the map editor and use the Publishing panel.
Click Export… on the target.
Pick a format and click Download.
The dialog lists every format in three groups: For people, For AI & agents, and For other systems.
Available formats
Group | Format | What you get | Plan |
|---|---|---|---|
For people | Print-ready document using the target's page layout | Every plan | |
For people | Word (.docx) | One | Business |
For people | Markdown | ZIP of | Every plan |
For AI & agents | Agent knowledge bundle (OKF) | ZIP of Markdown plus typed metadata on Google's open format | Business |
For AI & agents | RAG bundle | Flattened Markdown for retrieval pipelines | Business |
For AI & agents | llms-full.txt | Live endpoint on the published site | Every plan |
For AI & agents | Per-page Markdown (.md URLs) | Live endpoints on the published site | Every plan |
For AI & agents | Connect via MCP | The target ID an agent needs to query the content over MCP | Every plan |
For other systems | XLIFF (translation) | One | Business |
For other systems | DITA | ZIP of | Business |
For other systems | HDITA | ZIP of one | Business |
The three For AI & agents endpoint rows are URLs, not downloads. On a published web target they show Copy URL (or Copy target ID for MCP). On a target that is not published to the web they show Publish to web to enable.
If the content contains unfinished work, the dialog shows a banner before the format list: This export contains N draft paragraphs and M unresolved suggestions. Export is not blocked, so check the banner before sending files to anyone outside your team. Draft paragraphs are marked as drafts in PDF output.
What each format resolves
Formats fall into two camps, and the difference decides which one you want.
Lossless formats keep your structure. Conditions, variables, and component references survive as markers, so the content can come back into Topicary (or another structured tool) intact.
Resolved formats run the publishing pipeline first. Components are expanded, conditions are filtered against the target's condition profile, and variable tokens are replaced with values. What you get is one flat variant of the content, the same one a reader would see.
Format | Components | Conditions | Variables |
|---|---|---|---|
Markdown | Preserved as markers, plus a | Preserved | Preserved |
DITA | Preserved as conrefs, if the component has one (see below) | Preserved as profiling attributes | Preserved as keyrefs |
HDITA | Preserved as | Preserved as | Preserved as |
XLIFF | Expanded | Preserved, so translators see every variant | Preserved |
Word | Expanded | Filtered for the target's profile | Substituted |
OKF | Expanded | Filtered for the target's profile | Substituted |
Expanded | Filtered for the target's profile | Substituted |
Markdown
Downloads a .zip containing:
One
.mdfile for each topic, with YAML frontmatter carrying the title, in a folder hierarchy matching the map_components/with one.mdfile for each component the map references_meta/conditions.jsonwith your filter groups and their values_meta/variables.jsonwith the target's variable values
Nothing is flattened. Conditional blocks, variable tokens, topic links, and component references stay in the files as HTML-comment markers, which is what lets a re-import rebuild the structured content.
DITA
Downloads a .zip containing one .dita file for each topic and a .ditamap reflecting the map hierarchy. Topics become DITA concepts, tasks, or references according to their topic type. Filter groups become profiling attributes (audience, platform, product, or otherprops), and variables become <keyword keyref="..."> elements carrying the resolved value. Bookmap structure, glossary entries, and reltables are supported.
Component references only become conrefs when the component carries a DITA conref path, which is the case for components imported from DITA. Every other component reference is written as an XML comment naming the component. Either way the component's text is not in the file. Markdown is the only export that ships the component content itself, in _components/.
HDITA
Downloads a .zip containing one .html file for each topic: HTML5 with data-* attributes carrying the DITA semantics. There is no map file and no manifest.
Use HDITA when the receiving system reads HTML but you do not want to lose the structure. Conditions become data-props, variables become data-keyref, and component references become data-conref, so the content round-trips back into Topicary without flattening. The output is one DITA Open Toolkit hop away from DITA proper. As in DITA export, a component reference is a pointer: the component's own text is not in the zip. HDITA is available for a whole target, not for a single topic.
Word (.docx)
Downloads one .docx file with every topic as a section. Word export preserves headings, lists, tables, images, bold, italic, and code formatting. The content is resolved first, so what you get is a single flat variant.
OKF (agent knowledge bundle)
Downloads a .zip of Markdown pages paired with typed metadata, following Google's open Agent Knowledge Format, so agents and retrieval systems can consume the documentation without a custom parser. The content is resolved first. OKF is available for a whole target, not for a single topic.
XLIFF 2.0
Downloads an .xlf file with source and target segments for translation tools. Components are expanded, but conditional blocks and variable tokens are preserved in the segments, because translators need to see every variant they are translating. Import the completed files back through the translation panel.
Export a single topic
Open the topic in the editor.
Click the three-dot menu.
Select Export as Markdown.
The file downloads immediately as a single .md with YAML frontmatter. Component references are resolved first. Conditions and variables are not, because a single topic has no publication target to supply a condition profile or a variable set.
Markdown is the only single-topic export in the interface. Every other format runs against a target.
Some rich structures do not round-trip perfectly. Tabbed code groups, embedded videos, and custom callout styles have no direct equivalent in every format, so they are converted to the closest available representation. Review the exported files before importing them into another system.
See also
Publish a web site: publish as a live documentation site instead of exporting static files
Generate a PDF: another output format for offline distribution
Import formats: supported formats for importing content into Topicary
The publishing pipeline: how components, conditions, and variables are resolved before export
Topic types: topic types determine DITA element mapping during export