Import Obsidian vault
Upload a ZIP of your vault and Orkestra turns it into a Wiki page tree at the organization level, preserving folder hierarchy and rewriting internal wikilinks as Orkestra routes.
If what you want is to connect the vault to an AI agent without migrating the content, see the Obsidian + MCP guide. That pattern is for live reading from the agent; this importer is for a one-time migration of the content into the team's shared Wiki.
How to export the vault
Obsidian has no native export — the vault is a folder of markdown files. You just zip it:
- Close Obsidian (so nothing is writing during the zip).
- Find the vault folder in Finder/Explorer.
- Compress it to ZIP:
- macOS: right-click → Compress.
- Windows: right-click → Send to → Compressed folder.
- Linux:
zip -r vault.zip /path/to/vault.
- Size cap: 50 MB. If your vault is larger, exclude heavy folders (
attachments/, images) and re-zip. Orkestra doesn't import binaries anyway.
What gets imported
.mdfiles → Wiki pages (ORGANIZATION scope, PUBLIC visibility).- Folder hierarchy → parent-child relations in the wiki tree. Each folder creates an "index page" to group its children.
- YAML frontmatter — if the page has
title: "...", that value overrides the filename. The rest of the frontmatter is preserved inside the page body. - Wikilinks
[[Page]]→ rewritten as[Page](/wiki/<slug>)if the target exists in the vault. - Aliased wikilinks
[[Page|display text]]→[display text](/wiki/<slug>). - Broken wikilinks (target missing) → converted to plain text and counted in the preview.
- Embeds
![[file.png]]→ placeholder`[file.png (attachment not imported)]`. The reference stays visible so you know what's missing.
Out of scope
- Binaries (images, PDFs, audio, anything non-
.md) — not uploaded. Embeds become placeholders. .obsidian/folder — silently ignored (app config, not content).- Other dotfiles (
.DS_Store,.git/) — ignored. - Community plugins (Dataview, Templater, Kanban, Canvas, Excalidraw) — special files (
.canvas,.excalidraw) are ignored. Dataview queries stay as inert markdown code. - Tags (
#tag) — preserved literally in text but not mapped to Areas or labels. WikiPage has no nativetagsfield. - Folders deeper than 10 levels — pages inside flatten to the wiki tree root with a warning. This cap exists in the Orkestra wiki engine to prevent cycles.
- More than 2,000 pages total — the importer stops. For huge vaults, split by section.
- Bidirectional sync — doesn't exist. The import is a point-in-time snapshot. Edits after the import don't flow back to either side.
- File version history — only the current version is imported. Orkestra starts its own version history from there.
Requirements
- The import requires an active organization. You can't import into a personal workspace.
- Admin or wiki creator permissions in the org.
Warnings vs silent ignores
The preview surfaces warnings for: broken wikilink count, files exceeding the depth cap, non-markdown files found, individual file read failures. Dotfiles, plugins and binaries are silently ignored.
Steps in Orkestra
- Go to Settings → Import → Obsidian.
- Upload the vault ZIP.
- Review the preview: pages to create, broken wikilinks, files skipped by depth, warnings.
- Execute. Atomic commit.
Alternative: read the vault without importing
If you'd rather not copy content to the team Wiki and prefer to keep it in your local vault, you can wire it to your AI agent via MCP. See Obsidian + MCP.
Resources
- Import data — overview
- Obsidian + MCP — live-read pattern (doesn't copy content).
- obsidian.md