Lattice - Turn AI conversations into lasting intelligence.
by•
Lattice turns your ChatGPT, Claude, and Gemini chats into a visual, branching map of your thinking. It tracks every prompt and edit, stores everything locally, and lets you search, query, and fork your entire AI history — with citations — across all three.
Replies
Hey Product Hunt 👋
I use ChatGPT, Claude, and Gemini every day, and my best thinking kept getting buried in an endless scroll across all three. So I built Lattice.
It's a local-first Chrome extension that maps every conversation into a branching "Atlas," tracks your prompt edits as branches, and lets you ask questions across your entire AI history — "where did I go in circles on pricing?" — with clickable citations back to the exact message. You can fork a stuck thread into a different model with full context, and everything stays on your machine (your own API key, nothing uploaded).
Would love your feedback on what to map next.
Branching map view is genuinely useful, made it way easier to follow a long thread of prompts. Wish it had a dark mode.
@kyakl74353 Thanks Erdal, this genuinely made my launch day. The branching map is the whole reason I built Lattice, so hearing it made a long prompt thread easier to follow is exactly the signal I was hoping for. As for the dark mode, its going to come up in the next iterations of Lattice. I'll reply here when it ships.
One thing that would make this genuinely useful for me is letting me tag or group related branches into named projects, kind of like folders. Right now everything just sits in one long map and it gets messy fast when I'm juggling multiple research threads at once.
@ahmetdurma175 Really good callout, the "one long map gets messy with multiple threads" problem is real, and you've described it better than I have. Grouping branches into named collections (folders/projects) is on the roadmap. Quick one: would you want to group within a single conversation, or across different chats into one project? That changes how I'd build it.
Love the idea of visually mapping prompts across models. One thing that would make it really useful for me: let me export the whole branching tree as a Markdown outline or PDF so I can hand off the reasoning path to a teammate who wasn't in the original chat.
@bedriyestzn Hey Bedriye! Thank you for your thoughtful feedback, and that is a sharp suggestion. Full-tree export (Markdown outline + PDF) is exactly the direction I'm thinking to take Lattice in. Would a Markdown outline cover your handoffs, or is PDF the format your team actually passes around?
The branching map idea is genuinely useful for untangling long research threads. One thing that would make it stick for me though: a way to tag or annotate specific nodes with my own notes, so when I come back weeks later I can see why I took a particular fork instead of just the raw prompt trail.
@berilelyptz Hey Beril! Thank you for your feedback! Per-node notes/annotations are on the roadmap. They'll be shipping along with the next iteration of Lattice.