Queue prompts when Claude hits its rate limit or while it's responding. They fire automatically the moment Claude is ready — keep your flow.
promptq injects a queue panel into claude.ai. Type your next prompts while Claude is still responding — or while you're rate limited. They fire automatically in sequence the moment Claude is ready. No more losing your train of thought or checking back manually.
Two triggers, one queue:
- Claude responding — queued prompts fire the instant the send arrow comes back
- Rate limit hit — queued prompts hold until the limit resets, then fire automatically
Features
- Strip docks above the composer bar — always visible, click to expand
- Type prompts in the queue panel, hit
+ Add - Reorder queue items with ↑ ↓ buttons
- Countdown timer shows exactly when session and weekly limits reset
- Auto-fire on reset (toggleable)
- Wait for response mode — sends each prompt only after Claude replies
- Desktop notification when queue finishes
- Light and dark mode
- Zero data collection — everything stays local on your device
-
Clone the repo:
git clone https://github.com/owlguardco/promptq.git
-
Open Chrome or Brave →
chrome://extensions -
Enable Developer mode (toggle, top right)
-
Click Load unpacked → select the
promptqfolder -
Go to claude.ai — the promptq strip appears above the composer bar
- Type a prompt in the queue panel textarea and hit + Add
- Add as many prompts as you want — they stack in order
- Hit Fire queue now to start, or leave Auto-fire on and it runs automatically
- If Claude is mid-response, it waits for the arrow to come back before sending the next
- If you hit the rate limit, the queue holds and fires the moment the limit resets
Interceptor (content/interceptor.js) — runs in the page context and hooks fetch before claude.ai's own code uses it. Reads limit data straight from Claude's own traffic: the /usage endpoint (exact session and weekly reset timestamps) and the live SSE message_limit events that stream during responses. This is the same technique Claude Counter uses, and it's far more accurate than parsing the rounded percentages shown in the UI.
Content script (content/content.js) — the core. Receives limit data from the interceptor and watches the send button state (orange arrow = ready, stop square = streaming). Runs a 4-state machine: IDLE → STREAMING → LIMITED → FIRING. Injects the queue UI above the composer bar.
Background service worker (background/service-worker.js) — sets a chrome.alarms timer for the rate limit reset time. Fires even when the tab is backgrounded. Sends a LIMIT_RESET message to the content script when the alarm triggers.
Popup (popup/popup.html) — quick status and settings from the extension icon.
promptq reads Claude's usage endpoints to know when limits reset. These endpoints power Claude's own UI but aren't officially documented as public, so a claude.ai redesign can temporarily break limit detection until the extension is updated. This is a risk shared by every Claude usage-tracker extension. When it happens, the queue still works — you can always fire it manually with Fire queue now. Updates are pushed here; run git pull and reload the extension to get fixes.
All data stays local. The extension reads only claude.ai's own traffic and your lastActiveOrg cookie (to query the usage endpoint). It makes no external requests and collects nothing.
PRs welcome. Keep it focused — one tool, one job.
MIT