Example page made by Ghostlaunch. This product doesn't exist yet.
Back to the exampleChrome extension for Claude & ChatGPT
Threadkeep logs what you decided and ruled out during long Claude and ChatGPT sessions, warns when the chat starts drifting, and writes the handoff for your next chat.
Free for 3 handoffs a week. Chats stay in your browser.
The first hour is sharp. Then the chat starts losing the thread, and you either push through a degrading session or pay the re-explanation tax.
“An hour in, it suggested the library I already said no to.”
“I can't tell it's running out of room until it's already garbage.”
“Every new chat costs me 20 minutes re-explaining my stack.”
“My handoff prompt always drops the one decision that mattered.”
“I keep a decisions file just to paste into new chats.”
It runs alongside the Claude or ChatGPT tab you already use. No new chat app, no copy-paste routine.
Add Threadkeep to Chrome and open Claude or ChatGPT in the web app. There's nothing to connect and no account needed to start.
As the chat goes, Threadkeep keeps a ledger of what's decided, what's rejected and what's still open, captured while the chat is still sharp.
It flags when the context is getting full or a reply contradicts your ledger, so you move before the session degrades.
Edit the ledger, then paste the handoff into a new chat. It starts with your decisions, rejections and open questions in place.
A generic summary tells the next chat what you talked about. Threadkeep tells it what you decided.
Threadkeep tracks how full the conversation is and checks new replies against your ledger. When one brings back a rejected option or the context nears its limit, you get a warning.
Every item is tagged decided, rejected or open, with the reason in one line. The next chat knows not to suggest Mongo again, and why you dropped it.
One click turns the ledger into a structured handoff. Paste it into a fresh chat and skip the 20-minute re-explanation.
Chat content and ledgers are stored locally in Chrome. Nothing is sent to Threadkeep's servers, so your code and project details stay on your machine.
Built for you if
You run multi-hour sessions in the web app
You build or debug in Claude or ChatGPT for hours at a stretch, and the second hour is never as good as the first.
You're a solo dev or indie hacker
No team to remember decisions for you. The chat is your pair programmer, and it keeps forgetting.
You already have a handoff ritual
A summary prompt, a decisions file, a notes doc. Threadkeep does that job while you keep working.
Probably not for you if
A handoff prompt runs once the chat is already drifting. Projects and memory hold static files, not today's decisions.
| Criteria | Threadkeep | Handoff prompt | Projects & memory |
|---|---|---|---|
| Warns before the chat degrades | Yes | No | No |
| Keeps rejected options and why | Yes | Often dropped | No |
| Captures today's live decisions | Yes | If run in time | No |
| Handoff you can review and edit | Yes | Partly | No |
| Nothing extra to remember | Yes | No | Yes |
Start free. Upgrade when you're handing off more than a few times a week.
For occasional long sessions.
Free
For daily multi-hour build and debug sessions.
$7/ month
Both plans keep everything local in your browser. Cancel Pro anytime.
A prompt runs when you remember it, usually after the chat has already drifted, so the summary inherits the drift and tends to drop rejected options. Threadkeep logs decisions as they happen and warns you before things degrade. If 3 handoffs a week is enough, the free plan covers you.
No. Chat content and decision ledgers stay local in your browser. Nothing is sent to Threadkeep's servers.
At launch: Chrome only, with the Claude and ChatGPT web apps. It doesn't work in desktop apps, Cursor or Claude Code.
Projects and memory hold static files and preferences. They don't track the decisions you made in today's session, and they don't tell you when the current chat is losing the thread. Threadkeep does both, and works alongside them.
It estimates how full the conversation is and checks new replies against your ledger, flagging when one brings back a rejected option or contradicts a decision. It's a warning, not a verdict. You decide when to switch.
The ledger is editable. Fix, add or delete items before you generate the handoff, so the fresh chat gets exactly what you approve.
Install the extension and open a Claude or ChatGPT chat. There's nothing to configure to get your first ledger.
Get warned before your chat drifts, and start the next one with every decision and rejected option in place.