It’s a horrible feeling when you are working away with Claude or ChatGPT and then you get a message that you’ve reached your limit. You’re options are to buy more, and that can get expensive. You can blame the platform, like I used to. Hey, who hasn’t yelled at Claude a few times when you run out of credits. But alas I should of been blaming myself and not the platform!
I did some research and realized that Claude doesn’t count your messages. It counts tokens and most of us are hemoragging tokens without realizing it. I was for sure!
So I started to track my own usage, tested different approaches and came up with a list of habits that will save you a lot of tokens and frustration. Serenity now!
Here’s the deal.
1. Edit Your Prompt. Don’t Send a Follow-Up.
This is the one that got me.
When Claude doesn’t nail it on the first try, your instinct is to fire off a correction. Something like “No, that’s not what I meant” or “Try again, but this time do X.”
Don’t do this!
Here’s why. Every message you send gets added to the conversation history. And Claude re-reads ALL of it on every single turn. So you’re not just paying for your new message. You’re paying for every message that came before it. Again.
The math adds up qucik. At around 500 tokens per exchange, 5 messages costs you about 7,500 tokens. 10 messages? 27,500. By message 30, you’re burning 31 times more tokens than message 1 cost you.
The fix is simple. Click edit on your original message (my go to!), fix it, and regenerate. The old exchange gets replaced, not stacked. You’re rewriting history instead of piling onto it.
2. Start a Fresh Chat Every 15 to 20 Messages
This follows directly from the first point. Token costs don’t just grow. They compound.
A chat with 100+ messages at around 500 tokens per exchange? That’s over 2.5 million tokens burned. And the vast majority of those tokens are just Claude re-reading old history. One developer tracked his usage and found that 98.5% of tokens were spent on re-reading the conversation. Only 1.5% went toward actually generating the response. Crazy, right?
Think about that. Almost all your tokens go toward re-reading, not creating. I was doing this a lot!
My approach: when a chat gets long, I now ask Claude to summarize everything we’ve covered. Then I copy that summary, start a new chat, and paste it in as my first message.
3. Batch Your Questions Into One Message
A lot of people think splitting questions into separate messages gets better results. Almost always, the opposite is true.
Three separate prompts means three full context loads. One prompt with three tasks means one context load. You save tokens twice: fewer reloads, and you stay further from hitting your limit.
Instead of sending “Summarize this article,” then “List the main points” then “Suggest a headline,” just write: “Summarize this article, list the main points, and suggest a headline.”
Bonus: the answers are often better because Claude sees the full picture upfront. It knows where you’re headed. This ones a winner.
4. Upload Recurring Files to Projects
If you’re uploading the same PDF to multiple chats, Claude re-tokenizes that document every single time. Every. Single. Time. I wish I knew this from the start!
Use the Projects feature instead. Upload your file once. It gets cached. Every new conversation inside that project references it without burning tokens again.
If you work with contracts, style guides, briefs, or any document you reference regularly, this alone could cut your token spend dramatically. I use it for everything, and it helps my tokens last longer.
5. Set Up Memory and User Preferences
Every new chat without saved context wastes 3 to 5 messages on setup. “I’m a marketer, I write in a casual style, I prefer short paragraphs…” You’ve probably seen people start every prompt with “Act as a…” That’s tokens burned on repeat.
Go to Settings, then Memory and User Settings. Save your role, communication style, and preferences once. Claude automatically applies them to every new chat.
I did this months ago, and it eliminated an entire category of wasted tokens from my workflow.
6. Turn Off Features You’re Not Actively Using
Web search, connectors, and Explore mode. All of these add tokens to every response, even when you don’t need them.
Writing your own content? Turn off Search and Tools. Extended Thinking also consumes tokens. Keep it off by default. Only switch it on when your first attempt wasn’t cutting it.
Simple rule: if you didn’t turn it on intentionally, turn it off.
7. Use Haiku (on Claude) for Simple Tasks
Grammar checking. Brainstorming. Formatting. Quick translations. Short answers. Haiku handles all of this at a fraction of the cost of Sonnet or Opus.
Choosing the right model might be the most important decision you make every day with Claude. Haiku for drafts and simple tasks frees up 50 to 70% of your budget for the work that actually requires the heavy models.
My mental model is straightforward. Haiku for quick tasks. Sonnet for real work. Opus for deep thinking. You don’t drive a semi-truck to pick up groceries.
8. Spread Your Work Across the Day
Claude uses a rolling 5-hour window. It doesn’t reset at midnight. Your limit gradually decreases as older messages fall off.
So if you burn through everything in a single morning session, you’re leaving most of your daily capacity unused.
I break my day into 2 to 3 sessions. Morning. Afternoon. Evening. By the time I come back, my earlier usage has rolled off, and I’ve got a fresh allocation waiting.
9. Work During Off-Peak Hours
Starting March 26, 2026, Anthropic now uses your session limit more quickly during peak hours. That’s 5:00 AM to 11:00 AM Pacific on weekdays. Same query, same chat, but during peak hours, it hits your limit harder.
Your weekly limit stays the same. But how it gets distributed has changed. Running big tasks in the evening or on weekends stretches your plan significantly.
If you’re outside the U.S., peak hours might actually fall during your afternoon, so check your time zone against Pacific Time.
10. Enable Extra Usage as a Safety Net
If you’re on the Pro, Max 5x, or Max 20x plans, go to Settings then Usage and enable the Overage feature.
When your session limit is reached, Claude won’t lock you out. It switches to pay-as-you-go billing at API rates. You set a monthly spending cap so there’s no surprise bill, and sometimes Claude even throws you free credits.
This isn’t about saving tokens but you don’t want to lose momentum at the worst possible moment. Been there, done that!
Serenity Now!
All of this felt like a lot when I first started paying attention to it. But once these habits became automatic, I almost never hit my limits anymore.
Some people even find they can drop down a plan tier because they’re finally using tokens efficiently.
Claude doesn’t count messages. It counts tokens. The sooner you internalize that, the better your experience gets.
Read my next article…it’s a banger! If AI Took Your Job, Here’s How to Use It to Get the Next One