The problem
It's Wednesday night, the kids are finally down, and you're halfway through the welcome email for a new client.
Claude stops and tells you you've hit a limit.
You don't know if you broke something or if this is the part where you're supposed to pay.
What you need
- Claude on the Free or Pro plan (everything here works on Free; the two paid-only bits say so)
- The chat that stopped on you, or the one you use the most
- 10 minutes
Steps
1. Read the message, because it tells you which wall you hit. Claude has two different limits and they get confused all the time. If you see "5-hour limit reached - resets [time]," that's your usage limit: how much you can use Claude across all your chats over a stretch of time. If you see "Your message will exceed the length limit for this chat. Try attaching fewer or smaller files or starting a new conversation," that's the length limit: how long one single chat can get. Two different problems, two different fixes, and neither one means you did anything wrong.
2. Hit the usage limit? Wait for the reset time it shows you. On Free, the limit is session-based and resets every five hours, and the number of messages you get varies with demand, so there's no fixed count to plan around. On Pro ($20 a month, or $17 a month billed yearly), you get at least five times the usage per 5-hour session that Free gives you, plus a weekly limit that resets at a fixed time assigned to your account. If you're on a paid plan, open Settings > Usage to see progress bars for your five-hour session and your week. Paid plans can also buy usage credits if the thing truly can't wait.
3. Hit the length limit? Start a new chat, and here's why that works. This is where tokens come in. Claude doesn't read your words one by one, it reads tokens, little chunks of text. Anthropic's docs put one token at about 3.5 English characters, and that number changes with the language, so ojo if you write in Spanish or switch back and forth. The context window is Claude's working memory for one chat, and as the conversation goes, every message you send and every answer you get piles up in there, along with every file you attached. That pile is what hits the length limit. And bigger isn't better: Anthropic's own docs say accuracy and recall drop as the token count grows. A fresh chat starts with an empty pile.
4. Check one setting so long chats can keep going on their own. Go to Settings > Capabilities and make sure Code execution and file creation is on. It's on by default for Free, Pro and Max, but if you ever turned it off, turn it back on. With it on, when a chat gets close to its limit, Claude summarizes the earlier messages so the conversation can keep going, and you might see it "organizing its thoughts." A summary is shorter than the original, so in practice, if a price or a date from early in the chat really matters, say it again.
5. Stop re-pasting the same stuff. Put it in a Project. Your price sheet, your policies, your bio: anything you use over and over. Anthropic says documents you upload to a Project are cached, and cached portions count less against your limits than new content. Projects also load only the relevant parts into the chat instead of the whole file. Free includes up to 5 Projects. Keep the Project instructions short and remove files you're not using anymore, because those are on Anthropic's own list of ways to save space.
6. Turn off what this task doesn't need. Research and web search are on Anthropic's list of things that use more, along with long messages, big attachments and higher effort levels. So for a routine email, turn off extended thinking, pick a lower effort level, turn off connected apps the chat won't use, and tell Claude not to search the web when you don't need anything current. (On the new Claude rolling out since September 16, 2026, there's no web search switch: Claude searches when it thinks it helps, so saying "don't search the web" is how you turn it off.)
7. One task, one chat, one complete message. This part is practice, not an Anthropic rule, and it's built on what Anthropic does say: chat length counts against your usage, related questions should go together in one message, and a clear, specific message saves you the back-and-forth. Five little "oh and also" messages cost more than one good one. The prompt below is that one good message.
The prompt
Copy this into a new chat (inside your Project, if it's about your business). Fill in the brackets, delete any line you don't need, read it once before you hit send.
New task: [WHAT YOU NEED, IN ONE LINE].
What you need to know:
- [FACT 1: a price, a name, a date, a rule]
- [FACT 2]
- [FACT 3, or delete this line]
What I want back: [THE FORMAT, e.g. one email under 120 words, 3 options, a short table].
Tone: [e.g. warm and direct, no exclamation points].
Don't search the web for this. Everything you need is above.
If something important is missing, ask me all your questions in one message before you start.What it looks like filled in
Primer Paso Coaching keeps getting the same DM: "do you have a discount?" Her prices already live in her Project, but she spells out the ones this reply needs so nothing gets mixed up.
You
New task: a short, kind reply to a DM asking if I have a discount.
What you need to know:
- Single session is $150. The 5-session package is $600.
- The package has a payment plan: 2 payments of $300.
- I don't do discounts. The payment plan is how I make it easier.
What I want back: one reply under 70 words that I can paste into Instagram DMs.
Tone: warm and direct, no exclamation points, no apologizing for the price.
Don't search the web for this. Everything you need is above.
If something important is missing, ask me all your questions in one message before you start.What came back, in one message, no follow-ups needed:
Claude
Thank you for asking. I don't offer discounts, but I do make it easier to start: the 5-session package is $600, and you can split it into 2 payments of $300. If you'd like to try one session first, a single session is $150. Want me to send you the link for whichever one fits?
One task, one chat, short in and short out. She saved the prompt in her notes so the next DM takes thirty seconds.
Takeaway
When Claude stops, it's almost always one of two walls, and the message tells you which: the usage limit means wait for the reset, the length limit means start fresh. Everything else in this guide is about hitting those walls less, and most of it comes down to giving Claude less to carry. If you're in the middle of a long chat right now and can't afford to lose the thread, the lesson "When you hit your limit" in the full course, Learn AI, Build Your Business, has the prompt that moves it into a new chat without losing anything, plus how to decide when it's actually time to move up a plan.
Reply and tell me: which wall do you hit more, the 5-hour one or the "this chat is too long" one?
The next step
Read next · For everyone
Find a connector or skill and check it's safe before you add it, in 15 minutes
A connector or skill you checked yourself before you let it into your Claude.
New to Claude? Start Here in the free classroom. The free plan works.
Sources
All checked October 1, 2026.
- How do usage and length limits work? (Claude Help Center): usage vs length limits, what affects usage, automatic context management, Projects/RAG, shorter instructions, extended thinking, effort level, turning off tools and web search, usage credits
- Usage limit best practices (Claude Help Center): factors that use more, batching, being specific, Project caching, Settings > Usage on paid plans
- Troubleshoot Claude error messages (Claude Help Center): exact limit messages
- Get started with Claude (Claude Help Center): Free plan five-hour reset, messages vary with demand
- What is the Pro plan? (Claude Help Center): $20/month, five-hour session reset, weekly limit
- Claude pricing: $17/month billed yearly, "at least 5x more usage per 5-hour session than Free," Free includes up to 5 Projects
- Create and edit files with Claude (Claude Help Center): Settings > Capabilities, "Code execution and file creation," on by default
- Glossary (Claude docs): tokens (~3.5 English characters), context window as working memory
- Context windows (Claude docs): messages and responses accumulate, context rot