Burning through your Claude credits

If you’ve ever hit Claude’s usage limit mid-task and had to stop dead in your tracks, you’re not alone – it’s probably the question I get asked most in my community, at least once a week. And almost every time, the fix isn’t upgrading to a more expensive plan. It’s changing a few habits.

Here’s everything you need to know.

First, understand how Claude actually counts usage

Most people assume Claude counts messages. It doesn’t. It counts tokens – which is basically a measure of how much text is being processed, including everything you’ve typed, everything Claude has responded with, and every file you’ve uploaded.

Here’s the part that catches people out: every time you send a new message in a long conversation, Claude re-reads the entire chat history before it even starts on your new question. So by message 20, you’re burning through tokens at a rate that has nothing to do with the complexity of what you just asked. A 20-message session uses around 105,000 tokens. A 30-message session? Around 232,000. It compounds fast.

The other thing worth knowing is that your limit runs on a rolling 5-hour window – not a daily reset. That means if you blast through everything in one big morning session, you’re leaving most of your day’s capacity on the table. Spreading your Claude work into two or three sessions across the day is a simple habit that gives you significantly more to work with.

The biggest lever you’re probably ignoring: your model choice

This is where most people are quietly burning through credits without realising it.

Claude has three models: Opus, Sonnet, and Haiku. Opus is the most powerful and the most expensive in terms of token usage. Sonnet is the everyday workhorse. Haiku is the lightweight option for quick, simple tasks. If you’re on Claude Pro and you haven’t manually selected your model, you might be defaulting to Opus for tasks that Sonnet handles just as well – which is most of them.

For around 80% of tasks, Sonnet delivers results that are practically indistinguishable from Opus, and it processes faster too. The difference in performance between the two models is actually tiny – we’re talking about 1 to 2 percentage points on the benchmarks that matter for everyday business use. Opus earns its place for genuinely complex work: large-scale technical projects, deep multi-file analysis, highly specialised research. Writing captions, cleaning up a ClickUp template, drafting an email, planning content? That’s Sonnet territory, every time.

A simple framework that works: default to Sonnet, switch to Opus only when Sonnet genuinely can’t get the job done. Most small business owners will barely ever need to make that switch.

Start new chats more than you think you should

This one feels counterintuitive but it’s one of the highest-impact changes you can make.

Long chats are expensive chats – not because the questions get harder, but because the re-read overhead compounds with every single message. The fix is to start a fresh chat every 15 to 20 messages, and carry forward only what Claude actually needs to keep going.

Before you close a long chat, ask Claude to summarise the key points, decisions, and next steps. Copy that summary, open a new chat, paste it in as your first message, and keep going. You get the continuity without the bloated token cost.

If you’re a Biz Rebelution member, this is exactly what the Chat Handoff skill is built for – it does the summary and handoff automatically so you’re not hunting back through the chat trying to work out what to grab. It’s one of those small things that makes a surprisingly big difference to how far your credits stretch.

Use Projects – they’re doing more work than you think

Claude Projects aren’t just a way to keep your chats organised. They’re a legitimate token-saving tool.

When you add documents, brand information, or recurring instructions to a Project’s knowledge base, that content is cached – meaning Claude can reference it without it eating into your conversation budget the same way a freshly uploaded file does. If you’re the kind of person who pastes their brand voice guide, business background, or standard instructions into every new chat, you’re spending tokens on the same content over and over again.

Set up a Project for your business. Add your brand voice, your offers, your target audience, any documents you reference regularly. Then start every chat from inside that Project instead of a blank conversation. The difference adds up quickly.

Anthropic’s own advice is to keep your project context files lean – under 2,000 words where possible. A massive knowledge base sounds useful but if the files are bloated, Claude is burning tokens just loading them before you’ve typed a word.

Batch your prompts, not your tasks

Three separate messages asking three related questions means Claude loads your full conversation history three separate times. One message with all three questions means it loads once.

This doesn’t mean cramming unrelated things together – it means being intentional about grouping related questions into a single prompt. “Summarise this, pull out the key points, and suggest a headline” is one message. Sending each of those as a separate follow-up is three.

Being specific also matters more than people realise. Vague prompts invite Claude to explore – and that exploration costs tokens. The more precisely you tell Claude what you need and what the output should look like, the less it wanders, and the less it costs you.

Watch what you’re uploading

File attachments – especially large PDFs and documents – add significantly to your token count. Only upload what Claude actually needs for the specific task you’re doing right now. If you’re asking Claude to write a caption, it doesn’t need your entire 40-page brand guide.

Also worth knowing: all your Claude usage across different surfaces (claude.ai, Claude Code, Claude Desktop) counts toward the same limit. If you’re working across multiple tools in a session, that’s all coming from the same pool.

Do you actually need to upgrade to Max?

Honest answer: probably not yet.

Max is worth it if you’re using Claude heavily every single day, across complex and time-intensive tasks, and you’re consistently hitting your limit even after implementing the habits above. For most small business owners, fixing the workflow comes first – and often that’s enough.

The gap between Pro at $20 a month and Max at $100 is significant. Before you spend it, spend a week applying the tips in this post. If you’re still hitting limits regularly after that, then the upgrade conversation makes sense. But upgrading without changing how you use Claude is like getting a bigger petrol tank without fixing the leak.

The short version

• Switch to Sonnet as your default model – save Opus for genuinely complex work
• Start fresh chats every 15 to 20 messages and use the handoff summary trick
• Set up a Project for your business and store recurring context there
• Batch related questions into single prompts
• Spread your sessions across the day instead of burning through everything in one go
• Upload only what’s needed for the task at hand
• Fix your workflow before deciding whether to upgrade

Claude Pro is more than enough for most small business owners when you know how to use it. The limit isn’t the problem – the habits are. Change those first.

Want to make your Claude sessions even more efficient? Inside The Biz Rebelution, you get access to the Chat Handoff skill plus a growing library of AI tools, templates, and resources built specifically for small business owners. Join us here.