Import
How to bring a Logseq graph into Cozy
A file-based Logseq graph is a folder of Markdown files, so you can bring it into Cozy by uploading those files in the browser or pushing them through the ingest API. Cozy has no native Logseq importer today. Wikilinks and tags carry over as text, while block references and embeds do not resolve as blocks.
What you are importing
A file-based Logseq graph stores each page as a Markdown file, usually under pages/, with daily notes under journals/. Cozy stores Markdown, text, JSON and CSV files, so the files themselves need no conversion step. The work is in deciding where they land and checking the links afterward.
Be clear about the limits first. Cozy has no native Logseq importer, and we have not verified a lossless migration of a large graph. Treat the steps below as a manual route, test it on a small slice, and keep your original graph folder untouched until you are satisfied.
If you use the newer Logseq DB version, the graph lives in a SQLite database rather than Markdown files. Per the Logseq repository, the DB version is still in beta and warns that data loss is possible. Export your pages to Markdown from Logseq before following this guide, and back up first.
Choose a route
| Route | Best for | Limits |
|---|---|---|
| Browser upload | A first test, or a few hundred notes | 1 MB per file, 25 files per selection |
| Ingest API | The whole graph, repeatable runs | 25 documents per request, 1.5 MB each |
| Your own agent | Cleanup and restructuring after import | Needs an MCP connection |
Both upload routes put documents into the built-in Inbox by default. That is a good place to review them before they become curated pages.
Route 1: upload in the browser
- Copy your graph folder somewhere safe and work on the copy.
- Sign in at cozy.cloud and open the workspace you want, such as My wiki.
- Select files from
pages/in batches of up to 25. - Repeat for
journals/if you want your daily notes.
This is slow for a big graph, but it is a good way to see how your notes look before you automate anything.
Route 2: push through the ingest API
The ingest endpoint accepts JSON with a list of documents. Each document has an id and content. Because a stable id makes a re-push update the same source, use the file path as the id so you can run the import again without creating duplicates.
curl -X POST "https://cozy.cloud/api/wiki/workspaces/$COZY_WORKSPACE/ingest" \
-H "Authorization: Bearer $COZY_TOKEN" \
-H "Content-Type: application/json" \
-d '{"documents":[{"id":"logseq/pages/Project Alpha.md","content":"# Project Alpha\n..."}]}'
You can also send a single raw Markdown file with ?id= on the URL. A small script that walks pages/ and journals/, builds batches of 25, and posts them is enough. Deleted sources are never re-imported, so if you remove something in Cozy on purpose, a later re-push will not bring it back.
Create the bearer credential under Agents. It needs the source-write scope. Credentials are shown once, so store the value in an environment variable straight away.
What carries over
| Logseq element | In Cozy |
|---|---|
| Page files | Markdown documents |
[[Page links]] | Wikilinks, resolved by path, stem, title or alias |
property:: value lines | Read as relations when the value is a wikilink, for example relation:: [[target]] |
| Tags | Frontmatter tags are read; inline hashtags stay as text |
| Journals | Plain files with date names |
| Block references and embeds | Stay as text, with no block-level addressing |
| Whiteboards, plugins, themes | Not part of the Markdown files |
Cozy builds its graph from Markdown conventions on every save. Links that resolve become graph edges, and links that do not resolve are marked broken. Links inside code are ignored. Read more in typed links and the Markdown wiki guide.
The block model is the real trade-off. Logseq lets you address and embed a single bullet. Cozy works with whole Markdown files, which suits agents that read and edit pages, but you give up block-level references. If those are central to how you think, say so to yourself before you move.
After the import
Imported sources are evidence, and Cozy treats them as untrusted data. Curated pages live in wiki/ and are capped at 90,000 bytes, while imported sources can be up to 1,500,000 bytes. Move the pages worth keeping into the wiki, then run a health check.
Cozy Health reports broken links, duplicate titles and aliases, orphans, pages missing a type or description, and oversized pages, then lists fixes in priority order. A graph built from years of Logseq notes will likely show many broken links at first, because journal pages and namespace-style names do not always match across files. Fix the ones that matter and ignore the rest. See wiki health for how scoring works.
Then connect your own agent through the MCP server so it can read, search and tidy the result. Cozy runs no model itself, so the quality of that cleanup depends on the agent you bring.
Where Logseq is the better choice
Logseq is open source under AGPL-3.0, runs fully local, and has block-level references, queries and a plugin ecosystem that Cozy does not match. If your workflow depends on those, keep Logseq and consider Cozy as a place for agents to read a Markdown copy. The Cozy and Logseq comparison goes into the differences in more detail.
Frequently asked questions
Does Cozy have a Logseq importer?
No. Cozy has no native Logseq importer. You upload the Markdown files in the browser or push them through the ingest API, then check the result with Health.
Will my block references and embeds still work?
No. They remain as text, because Cozy works with whole Markdown files and does not address individual blocks. Page links written as wikilinks can still resolve.
Can I run the import more than once?
Yes, if you give each document a stable id, such as its file path. A re-push then updates the same source instead of creating a copy. Sources you deleted are not re-imported.
What if I use the Logseq DB version?
Export your pages to Markdown from Logseq first, since Cozy reads files rather than the SQLite graph. Back up before exporting, as the DB version is still in beta. See import Markdown for the general route.
Give your agents a memory worth keeping
Cozy gives every account a private wiki that Claude, Codex, ChatGPT and other MCP clients can read and maintain.
Start your wiki