Save Claude chats as linked Logseq pages

To move a chat from Claude to Logseq, export Markdown and add it to a Logseq OG graph’s pages folder. In an existing database graph, start by pasting one turn into a new page. Organize the saved conversation into blocks and link it to your project and journal.

Check which Logseq version you use

As of November 2026, Logseq has two separate apps: Logseq OG for file graphs and Logseq for database graphs. The folder method below is for OG.

Both are outliners. If you prefer keeping the conversation as a conventional Markdown document, the Obsidian workflow covers that route.

Export Markdown into a file graph

DocMyChat saves the open claude.ai conversation as Markdown in Google Chrome on desktop. Markdown downloads are free.

1. Open the conversation in Chrome with DocMyChat installed.
2. Click Export next to Share.
3. Choose Markdown.
4. Give the downloaded file a descriptive name, such as workshop-feedback-chat.md.
5. Copy it into the pages folder inside your Logseq OG graph.
6. Open the page in Logseq OG.
7. Compare the first and last turns with the conversation.

If the page doesn’t appear, re-index the graph. Logseq’s file-import guidance describes this route for individual Markdown files. Keep the original download while you tidy the working copy.

Use the export icon under a Claude reply to save just that answer. Without an extension, select a short passage in the browser and paste it into a new page. Include the prompt and chat URL when the answer needs context.

The Markdown export guide explains the file’s metadata and artifact appendix.

Add the chat to a database graph

For one chat in an existing database graph, open the exported file in a text editor. Paste the first turn into a new page and check its blocks. Add the remaining turns, then tidy the page using the steps below.

If you’re migrating a whole file graph, add the chat to that graph first. Logseq’s File to DB graph importer is under Import in the upper-right three-dot menu and asks for a graph name. It converts a whole graph into a separate database graph. It doesn’t add one Markdown file to your existing database graph.

Turn the transcript into an outline

A full export uses ## You and ## Claude headings for the speakers. These are Markdown headings, not links to separate pages. Long answers may need splitting into smaller blocks.

Inspect one turn before rearranging the whole page. If several paragraphs sit in one block, split them where the subject changes. Keep quotations and code intact.

Logseq’s outliner uses indentation to group related blocks. Use Tab to indent a block beneath its parent and Shift + Tab to bring it back out. A speaker block can contain the answer’s explanation, example, and follow-up as children. Collapse it to scan the other turns.

For example, a workshop-planning chat could have a “Claude — proposed feedback questions” parent with three question blocks underneath. Keep your next correction under a separate “You” parent. This preserves who proposed each change while making the questions easier to reuse.

In the database version, heading formatting no longer exposes its Markdown markers. A heading without editable ## characters is expected there.

Check math, code, and tables

The export keeps formulas as LaTeX between dollar signs. Logseq OG’s Markdown reference shows math with double-dollar delimiters, such as $$E = mc^2$$. If an expression surrounded by single dollar signs stays literal, try that syntax in your working copy. Check the rendered expression, especially fractions and subscripts.

Code exports include fences and language names. Keep each snippet together when splitting blocks. For database migration, enable Extract inline code snippets as child blocks if a block contains several snippets. The importer otherwise keeps only the first snippet in that block.

For a pipe table, compare the header and every row with the original. If it appears as plain text, inspect the header separator and keep the rows together while repairing it. For a short comparison, separate labeled blocks may be easier to maintain.

Database table views work with graph data. Preserve the chat table’s meaning before deciding whether to rebuild it in that format.

Save essential images and files separately

Markdown image references point to claude.ai and need a signed-in browser session. They aren’t local image copies for offline reading.

Save an essential image separately and add the local file to your note. Keep its explanation beside it if later blocks refer to “the chart above.”

The artifacts ZIP contains the latest version of each artifact or file. It has no folders, so arrange extracted files alongside your notes.

Published Claude Docs and decks appear in the conversation export as titled links, without their contents. Slide HTML and JSON used only to build a published deck are left out of the appendix and ZIP.

Link the page to your project and journal

Use double-bracket page links to connect the conversation to existing work. Choose an existing page from the suggestions so a spelling difference doesn’t create a second project page.

For example, name the imported page “Workshop feedback chat.” Add a block saying “Source conversation for [[Customer workshop]].” Beneath it, write your decision: “Use three open questions and remove the satisfaction score.” Keep that decision separate from the transcript.

Then open today’s journal and add “Revised [[Customer workshop]] feedback questions after reviewing [[Workshop feedback chat]].” You can reach the source from the day’s work and the project without copying the full exchange into both places.

Keep the exported link back to Claude so you can return to the source conversation when it’s available.

Connect an AI tool separately

The database app’s optional MCP server lets compatible AI clients work with a graph. Follow Logseq’s setup guide if you need that connection. Saving a chat as described above doesn’t require it.

Questions