Collection Mode is Savelore's most powerful feature. It lets you pick specific extracts from multiple AI conversations — across different platforms, different chats, different days — and combine them into a single, organized note with citations that link back to each original message.

Think of it as building a Wikipedia-style article from your AI research, where every sentence traces back to a source you can verify.

What it is

Normal saving captures one full response. Collection Mode is different: you highlight specific passages from multiple responses across multiple conversations, and they get woven together into one note.

Each extracted passage keeps a citation — a numbered marker [1], [2], [3] — that links directly back to the original AI message where you found it. Click the citation and Savelore opens that conversation and scrolls to the exact message.

This makes Collection Mode ideal for research. You spend time across multiple ChatGPT and Claude sessions, collecting the best pieces of each response, and end up with a clean structured document instead of 12 scattered browser tabs.

How to activate Collection Mode

Collection Mode is in the Savelore sidebar toolbar — the row of icons at the top of the sidebar.

1. Open the Savelore sidebar by clicking the toolbar icon
2. Click the Collection Mode icon (stacked layers icon) in the sidebar toolbar
3. A collection panel opens at the bottom of the sidebar
4. Name your collection (e.g., "React performance research")

How to add extracts

With Collection Mode active:

1. Go to any AI conversation (ChatGPT, Claude, or Gemini)
2. Highlight any text in an AI response by clicking and dragging
3. A floating button appears: Add to Collection
4. Click it — the extract is added to your collection with a citation

Repeat this across as many conversations and platforms as you want. Each new extract gets the next citation number.

What the final note looks like

Your collection note contains the extracts in order, each followed by a citation marker:

React uses a virtual DOM to minimize expensive real DOM updates. When state
changes, React first updates the virtual DOM, then diffs it against the previous
version, and only applies the minimum necessary changes to the real DOM. [1]

For lists and repeated elements, always provide a unique key prop. React uses
keys to track which items changed, were added, or were removed. [2]

Memoization with useMemo and useCallback prevents child components from
re-rendering when their props haven't actually changed. [3]

Citations [1], [2], [3] are clickable links. Each one opens the original conversation and scrolls to the exact AI message the extract came from.

Real example: researching a topic across platforms

Suppose you're learning about database indexing. Over a few sessions you've had:

• A ChatGPT conversation explaining B-tree indexes
• A Claude conversation about index selectivity and query planning
• Another Claude conversation about when NOT to use indexes

Here's how you'd use Collection Mode:

1. Start a collection — name it "Database indexing"
2. Open the ChatGPT conversation — highlight the clearest explanation of how B-tree indexes work → Add to Collection [1]
3. Open the first Claude conversation — highlight the paragraph on query planner behavior → Add to Collection [2], then highlight the selectivity rule → Add to Collection [3]
4. Open the second Claude conversation — highlight the counter-examples (when indexes hurt performance) → Add to Collection [4]
5. Finish the collection — click "Done" in the sidebar panel
6. Export to Notion — your structured note lands in your Notion database, citations intact

You now have a single, coherent reference document built from four separate AI conversations across two platforms. Every claim is sourced and one click away from verification.

Exporting a collection

Collections can be:

• Exported to Notion (Pro) — sends the note to your connected Notion database with citations as inline links
• Downloaded as Markdown (Pro) — a .md file with citation footnotes
• Downloaded as HTML (Pro) — for importing into other tools
• Copied to clipboard — available on all plans

Tips for getting good collections

Be selective. Don't add every paragraph — just the specific passages that directly answer your question. A tighter collection is more useful than a comprehensive dump.

Use consistent naming. Name your collection before you start adding extracts. It's harder to rename later.

Work across platforms. Collection Mode's real power is stitching together Claude's explanations with ChatGPT's code examples. Don't limit yourself to one platform.

Citations are permanent. As long as you don't delete the original conversation, the citation links will always work.