Memory Sync
Take your AI memory with you across Claude, Gemini, and more
Start new thread
Zhengyang Hou

Memory Sync - Take your AI memory with you across Claude, Gemini, and more

by
YourAIScroll
•
Memory Sync helps you take your AI memory with you. If you use multiple AI assistants, you’ve probably had to repeat the same preferences, context, and instructions again and again. Memory Sync gives you one portable Memory.md so you can pull memory from one platform, edit it in one place, and push it into another. It currently supports ChatGPT, Claude, Gemini, Grok, Kimi, Mistral, and Copilot, and also tracks last sync times so you can see what’s up to date.

Add a comment

Replies

Best
Zhengyang Hou
Maker
📌
Hey Product Hunt 👋 I built Memory Sync because I was tired of reteaching every AI the same things about me. I use different assistants for different jobs, and the memory layer was getting fragmented fast. One knew my preferences, another knew some project context, another knew basically nothing. It felt messy and kind of dumb. So I wanted a simple, portable memory layer that I could actually control. The core idea is: • keep one Memory.md as the source of truth • pull memory from one platform • edit / clean it up • push it into another platform • track sync times so you can tell what’s current Right now Memory Sync supports ChatGPT, Claude, Gemini, Grok, Kimi, Mistral, and Copilot. One deliberate choice: it’s not fully automatic yet. The extension opens the target site, injects the sync prompt, and tracks sync state, but you still review and send things yourself. I wanted the first version to be transparent and controllable, not “magic” but hard to trust. Would love your feedback on a few things: • Is portable AI memory a real problem for you? • Should the long-term format stay Markdown-first? • Would you want more structure, or do you prefer keeping it simple? Website: https://mem-sync.kareverie.com/ Chrome extension: https://chromewebstore.google.co... Happy to answer any questions — especially tough ones.
Thami Benjelloun

Hello @zhengyang_hou ! Congratulations...
how do you prevent the Memory.md file from degrading over time once users start accumulating conflicting instructions across different assistants? is there any conflict resolution layer when platforms disagree on what the user is or prefers?

Zhengyang Hou
Maker

@thamibenjelloun Great question. I’m treating that as a curation problem, not a blind merge problem. Memory Sync is intentionally human-in-the-loop right now because I don’t think long-term memory should auto-merge silently across assistants. Different platforms can infer different things, and some of that will be noisy or contradictory.

So the current model is: pull memory out, review it in one editable Memory.md, compress / clean it up, then push forward only what should persist. In other words, conflict resolution currently happens through user review, not hidden automation. Longer term, I do think there’s room for better structure, conflict flags, and suggested deduplication, but I wanted the first version to optimize for transparency and control.