I’m a 60-year-old dad with no coding background. Before I try a new tool, I usually spend a day just reading about it. I didn’t do that with Claude Code. I opened it and started building right away, adding small amounts of credit whenever I ran low.

Add up all those small payments, and the first month cost about $75.

Two Doors, and I Walked Through the Wrong One

I didn’t know Anthropic has two completely separate ways to pay for Claude Code.

One is a subscription — the same kind of plan you’d use for claude.ai chat, at a flat monthly price. The other is the API, where you pay per token, based on how much you actually use.

I assumed there was only one option: the API, which is prepaid — you buy credit and Claude Code spends it per token. So that’s what I connected Claude Code to.

There was no single $75 bill. The API is prepaid. It runs out of credit, and you add more, a little at a time. Add up all those small top-ups, and the total for that month was about $75.

Before that, I had been using Cursor AI on its $20/month Individual plan. I hit its usage limit a few times and paid a bit extra to keep going — my total there usually landed around $30. Even with the overage, Cursor was cheaper than the mistake I made with Claude Code’s API billing.

The $20 Plan I Should Have Started With

Once I understood the two billing tracks, I signed up for Claude Pro at $20/month — the same subscription you’d use for the regular claude.ai chat — and pointed Claude Code at that subscription instead of an API key.

That’s when I learned Claude Code’s subscription plans don’t run on a simple monthly quota. They reset on two clocks at once:

  • Every 5 hours — a short-term usage window
  • Every 7 days — a longer weekly window

Both limits exist at the same time. As far as I can tell, you’re managing two different budgets, not one.

Where I Actually Check My Usage

I didn’t find this by reading documentation. I found it by clicking around.

In the Claude Code interface, there’s a small user icon in the bottom-left corner. Clicking it opens a settings menu, and inside that menu there’s a Usage section. That’s where the actual 5-hour and 7-day numbers live — how much of each window I’ve used, a bar showing the percentage left, and roughly when each one resets.

I check it in the middle of a work session now, not after the fact. If I see the 5-hour window getting close to the limit, I slow down or switch to a smaller task instead of pushing through.

Why the 5-Hour Reset Actually Fits My Day

I have a 3-year-old daughter, so my day is already cut into pieces by housework and childcare, whether Claude Code likes it or not. At first the 5-hour window sounded like a limit I’d have to fight. It turned out to fit the day I already have.

When I use up a 5-hour window in the middle of a big task, Claude Code tells me exactly when it resets, down to the minute. So I stop, go do whatever housework is next, and come back right when the window opens again. No guessing, no checking back every few minutes.

The percentage bar changes how I plan too. If I know chores are coming up soon anyway, I sometimes push through and spend most of what’s left in the current window on purpose, instead of saving it carefully — because I know a fresh window is coming the moment the chores are done.

How I Keep My Token Use Small

Watching the usage screen only goes so far. I also changed how I work to use fewer tokens per session in the first place.

I lowered the context memory setting so Claude Code warns me once a session’s context grows past a certain size, telling me it’s time to wrap up instead of pushing further into a long conversation. I try to keep each session to one small unit of work: finish the task, update CLAUDE.md, update memory, commit, run a backup skill for the important code files — then close the session or run /clear to reset the context completely.

That sounds like it should make the next session harder to start, since the memory is gone. It doesn’t, in practice. Between CLAUDE.md, the git commit history, and my saved coding-history notes, picking up where I left off takes no real effort. The next session reads those instead of carrying the old context forward.

The First Month Still Went Over

Switching to the $20 plan didn’t fix everything right away. I hadn’t built the habit of checking the usage screen yet, so I still went over the limit a few times that first month and paid for the extra usage on top of the subscription.

After that, I started checking the 5-hour and weekly numbers regularly, mid-session, instead of only when Claude Code told me I’d hit a wall. Since then, I’ve stayed inside the $20/month plan with no extra charges.

What I’d Tell a Beginner

If you’re just starting with Claude Code, figure out which billing track you’re actually on before your first real coding session, not after the small charges add up. If you’re on a subscription, find the usage screen early — bottom-left icon, then Settings, then Usage — and get in the habit of checking it while you work, not just when you get cut off.

$75 for one confused month taught me that. $20 a month is what it costs once you know where to look.