Claude Code pricing: what it really costs, plan by plan

Claude Code prices at a glance

Claude Code isn’t billed separately: it comes with your Claude subscription. You pay one monthly fee, and the chat and the terminal agent draw from the same usage allowance. Here are the plans, verified in June 2026:

• Pro: $17/month with annual billing, $20 monthly — Claude Code included.
• Max 5x: $100/month — 5 times the Pro plan’s usage capacity per session.
• Max 20x: $200/month — 20 times the Pro plan’s capacity per session.
• API: no subscription, you pay for the volume of text processed (see below).
• Claude’s free plan: chat access, but no Claude Code.

The Pro plan at $17–20: the right starting point

To discover Claude Code, the Pro plan is enough in the vast majority of cases. You get the regular Claude assistant (browser, app) plus Claude Code in the terminal, with usage limits shared between the two: chain heavy work sessions and you can hit the ceiling and have to wait for it to reset.

For non-developer work — writing, structuring files, keeping routines — those limits are rarely a problem at first. Practical advice: start with monthly billing at $20 to test for a month or two, then switch to annual at $17/month once you’re sure of your usage.

The Max plans at $100 and $200: for heavy users

The Max plans don’t unlock a different product: it’s the same Claude Code, with far more capacity. Max 5x ($100/month) gives 5 times the Pro plan’s usage per session; Max 20x ($200/month) gives 20 times.

You need it if you regularly bump into the Pro plan’s limits: long daily sessions, large volumes of files, several missions in parallel. If you’ve never seen a limit-reached message, you don’t need Max. The sound reflex is to upgrade when your usage demands it — not before.

The API: pay as you go, no subscription

Third option: connect Claude Code to an API key and pay for the volume of text processed, measured in tokens (fragments of words). As of June 2026, the Sonnet 4.6 model costs $3 per million tokens read and $15 per million tokens produced; Haiku 4.5, lighter, costs $1 and $5; the more powerful Opus models, $5 and $25.

On paper it looks cheap. In practice, Claude Code reads your files, thinks and chains steps: consumption climbs fast and the bill is hard to predict when you’re starting out. The API makes sense for occasional or automated use; for regular work, the fixed-price subscription is simpler and more predictable.

How to choose without getting it wrong

Three markers to decide. One: if you’re discovering the tool, take Pro with monthly billing — $20 committed, cancellable, more than enough to form an opinion. Two: only move to Max if the limits concretely block your work. Three: keep the API for special cases — one-off tests, integrations — not for getting started.

Two details that prevent surprises. Anthropic’s prices are listed in US dollars, before tax: depending on your country, VAT may be added. And offers change regularly — Team plans also exist for teams, from $20 per person per month with annual billing — so always check Anthropic’s official pricing page before committing.

Getting your money’s worth: the folder matters more than the plan

The real question isn’t only the price: it’s what you get out of the tool once subscribed. Claude Code reads the folder you open it in before working. Opened in an empty folder, even the $200 plan answers like a generic chat; opened in a folder that describes your job, your rules and your missions, the $20 plan already produces useful work.

That’s exactly what the Claude Code AI Workspace Kit prepares: you answer a few questions about your line of work, and your Context Folder is generated — an AGENTS.md defining the AI’s role, your rules and typical missions. The kit is being built: sign up on its page to be told when it ships.