Guide

How much does Claude Code cost?

Updated August 2026 5 min read

Claude Code doesn't have its own price tag — it comes with a Claude subscription, or runs against the pay-per-token API. Which one is cheaper depends less on the sticker price than on how you work. Here's the honest breakdown, as of August 2026.

The short answer

As of August 2026

$20/month (Pro) gets you Claude Code with an entry-level usage budget. $100/month (Max 5x) and $200/month (Max 20x) buy the same tool with 5× and 20× the budget. Or skip subscriptions and pay per token on the API — no ceiling, no floor.

Every subscription tier meters usage the same way: a rolling window of a few hours plus a weekly cap, both of which scale with the plan. The mechanics are covered in our Claude Code rate limits guide — what matters here is that the price buys headroom, not features.

Subscription plans, and what more money buys

Pro — $20/month

The entry point, and enough for regular but moderate use: focused sessions, smaller tasks, one agent at a time. A couple of long autonomous runs in a day can hit the rolling window — if that keeps happening, that's the upgrade signal.

Max 5x — $100/month

Five times the Pro budget. This is the tier where daily agentic work — long refactors, parallel sessions — stops bumping into the window on ordinary days.

Max 20x — $200/month

Twenty times the Pro budget, for people who run Claude Code like a full-time teammate — multiple parallel agents, long autonomous sessions, all day. Even here the weekly cap exists; it's just much further away.

Providers adjust caps over time, so treat multipliers as the durable fact and any specific message count you read elsewhere as a snapshot. The honest way to know whether your tier fits is to watch your own window consumption for a week.

The API alternative

Claude Code can also run against an Anthropic API key instead of a subscription. You pay per million tokens — as of August 2026, Opus-tier models are $5 per million input tokens and $25 per million output, and Sonnet-tier is $3/$15. There's no window and no weekly cap; the meter just runs.

The API wins when your usage is light or spiky — a few sessions a week, CI jobs, occasional bursts. The subscription wins when usage is heavy and constant: long agentic sessions consume tokens far faster than most people expect, and a single ambitious day on the API can cost more than a month of Pro.

What actually determines your cost

The plan sets the ceiling, but your working style sets the spend. Autonomous work is the multiplier: every file the agent reads, every test it runs, every retry counts against the budget — so two developers on the same plan can have wildly different headroom left by Friday. That's why the practical cost question isn't "which plan is cheapest" but "how fast am I burning the one I have" — which you can't answer without seeing the meter.

See what you're spending as you go

Vibe Island reads your existing Claude Code login and shows the live state of each usage window in your Mac's notch — percentage used and time until reset, updating as you work. No command, no browser tab. When you're deciding whether $20 is enough or $100 is worth it, a week of glancing at real numbers answers it better than any pricing table.

5h 11% 4h1m 7d 2% 6h1m

A real crop of the Vibe Island notch — each provider's windows as live, color-coded numbers, so you always know how much budget is left.

Common questions

Can I use Claude Code on the $20 Pro plan?

Yes. Claude Code is included with the Pro plan; the Max plans buy the same tool with a much larger usage budget. On Pro, a few long autonomous sessions in one day can reach the rolling-window limit, which is the usual reason people upgrade.

Do Claude Code and the Claude chat app share the same limit?

Yes — a subscription has one usage budget, and Claude Code sessions and chat conversations draw from the same rolling window and weekly cap. Heavy coding days leave less room for chat, and vice versa.

Is the API cheaper than a Max subscription?

It depends entirely on volume. The API bills per token with no ceiling, so light or spiky usage often costs less than $100 a month — and heavy daily agentic use usually costs far more. The subscription is effectively a flat rate for a large but capped budget; the API is a meter that never says no.

How do I know how much of my limit I've used?

Inside Claude Code, /usage shows your account's current windows. If you'd rather not ask a CLI every time, Vibe Island shows the live percentage used and time until reset for each window in your Mac's notch, updating as you work.