// the 10-second version
A fuel gauge for your AI chats. It shows how much fuel is left, warns you before you run out, and packs your conversation into a suitcase so you can carry on with a different AI.
// why it's cool
Limits shouldn't break your flow
AI usage is measured in tokens, chunks of text roughly three-quarters of a word long. Run out mid-session and all the context you built up gets stuck. FlowKeep makes the limit visible and turns hitting it into a handoff instead of a dead end.
// how it works
Four features
- Live gaugeA draggable badge on claude.ai with a colour-coded usage bar and a reset countdown.
- Pre-flight checkWarns you while typing if a message will likely blow your remaining budget.
- Auto snapshotAt 80% usage, saves a compact summary: your goal, recent context, and how to pick up again.
- One-click handoffCopies the snapshot and opens Gemini, ChatGPT or Groq so you can paste and keep going.
// under the hood
Counting without a counter
The browser can't see exactly how the model splits text into tokens, so FlowKeep estimates it from character and word counts, adjusted for code, which uses more tokens. When Claude's own responses include real usage numbers, it uses those instead.
It's about 650 lines of plain JavaScript on Manifest V3, the current extension standard. One codebase runs on Chrome, Edge, Brave and Firefox, and everything stays in your browser.
// the tricky part
Packing the suitcase
Paste the whole conversation into a new model and you blow its limits too. Paste too little and it's lost. The snapshot keeps just the goal, the latest context and a clear "continue from here", small enough to paste anywhere and complete enough to keep going.