New to Claude Cowork and hitting your session limits too quickly? Start here
🍊 TL;DR: Two Rules for better prompting & longer sessions + commentary
No extended back-and-forths with your AI. When the conversation includes more than three prompts that's the cue to ask for a summary and start a new session (alternative to run /compact). It's the single best thing you can do to stay inside your session limit and keep working. Why long conversations get expensive · Should I start fresh?
Pack as much context as possible into one prompt. If using Claude, use Plan Mode extensively — don't plan → code → test as three separate prompts. Gather them into one rich ask. How to structure prompts
My goal is to help you extend your AI sessions. There's a growing tendency to treat token usage as a proxy for productivity, where more tokens = more value delivered. That's a false equivalence. To me, maxing your token usage means working within the system and being a smarter prompter. Most people don't have unlimited tokens...get the most from the ones you have.
🔬 Prompt Analyzer
How token-efficient is your prompting?
Paste your prompt below. We'll estimate the token cost and suggest a better version - no AI involved!
0 characters
🔒 Your prompt is analyzed entirely in your browser and is never saved, stored, or transmitted anywhere.
Input Tokens
—
Est. Output Tokens
—
estimated
Total Estimated
—
input + output
⚡ Session overhead— added to every message on top of the prompt cost above
Active MCP servers
0
× ~4,000 tokens each
CLAUDE.md / memory files~600 tokens
Skills / system tools~500 tokens
Number of Prompts in Conversation with AI
1
first message — no prior context
Overhead per message~1,100 tokens
⏱ 5-Hour Session Usage— cumulative tokens consumed across all messages this session using Claude as an example
Estimated cumulative usage—
Budget remaining—
0%
0%25%50%75%100%
💡 Start a new chat (in Chat or Cowork) to reset your token usage. The 5-hour usage window is rolling — it starts with your first message and resets when you send a new message after the window lapses.
✨ Optimized Prompt
Your voice is preserved — only minor structural tweaks and filler word removal were applied. Don't stress about phrasing things perfectly; the AI handles that just fine. What actually drains your session is multiple prompts on the same topic. Pack more context into one prompt instead of following up — one rich ask beats three short ones every time.
Token savings:
Have questions or feedback about these results? Contact me
📈 Why Every Message Costs More
Token cost compounds with every reply
Claude re-reads the entire conversation on every message. A 30-message chat balloons to 232K+ cumulative tokens.
🎓 Token Education
Smarter Prompting, Longer Sessions
Everything you need to know about squeezing more out of every Cowork session.
🔒
Your prompts never leave your browser
No exceptions. No fine print.
✓
No AI processes your prompt. Every estimate and optimization is pure JavaScript running locally in your tab.
✓
No network requests are made. Open DevTools → Network tab. You'll see zero outbound calls when you analyze a prompt.
✓
Nothing is stored or logged. Close the tab and everything is gone. There is no database, no backend, no analytics on your input.
✓
Works offline. Once the page loads, it needs no internet connection to analyze your prompts. The logic is entirely local.