Is Claude Code free?
Claude Code is included at no extra charge in Anthropic's Pro and Max subscriptions, so on those plans there is no per-use fee, only usage limits. On an Anthropic API key you pay per token instead.
Guide · Cost5 min read
What Claude Code actually costs (the real token math)
There is no single Claude Code price, because it depends on how you pay. On a Pro or Max subscription it is a flat fee with limits. On an Anthropic API key you pay per token, and the bill scales with the model you pick and the tokens you burn. Here is how each works, why real-world bills have shocked whole teams, and how to know your own number before the invoice does.
Quick answer
Claude Code is included in Anthropic's Pro plan (about $20/month, or roughly $17 billed annually) and Max plans ($100/month for 5x usage, $200/month for 20x), where the cost is flat and limits are the real constraint. On an Anthropic API key you pay per token instead, so model choice dominates: Opus output is billed around $25 per million tokens, and thinking tokens bill at output rates too. Anthropic has cited an average of roughly $150 to $250 per developer per month for Claude Code, with power users far higher. Tokens 4 Breakfast reads your local session token counts and shows usage, per-model spend, and a month-end forecast from the Mac menu bar.
Step by step
The full walkthrough
Each step stands on its own — skip to the one that matches where you are.
Know which of the two ways you are paying
Claude Code bills one of two ways. On a Pro or Max subscription you pay a flat monthly fee and get usage limits. On an Anthropic API key you pay per token with no flat cap. Some people use both. The real cost question is completely different depending on which one you are on.
On a subscription, cost is fixed and limits are the constraint
Pro is about $20 per month (roughly $17 billed annually), and Max is $100 (5x usage) or $200 (20x). The dollar figure does not move, so what actually bites is hitting the rolling 5-hour window or the weekly cap mid-task. On a subscription, track usage against your limit, not an invoice.
On an API key, model and tokens drive the whole bill
Pay-per-token means the model you pick dominates. Opus output runs around $25 per million tokens, Sonnet is far cheaper, and cache reads are cheap. Thinking tokens bill at output rates, so high reasoning effort multiplies cost. A heavy Opus session with large context can cost many times a lean Sonnet one for similar work.
Understand why real bills shock people
Coverage in 2026 has shown teams blindsided by usage-based reality: reports of roughly $2,000 per engineer per month on Claude Code, individual write-ups of $1,600 bills, and at least one company burning its entire annual AI-coding budget months early. The pattern is always the same: nobody watched the per-session burn until the invoice arrived.
Find your own number before the invoice does
The only cost that matters is yours. Tokens 4 Breakfast reads your local Claude Code session logs, shows tokens and estimated spend by project and model, and projects your month-end total, live in the menu bar with no API key required. That turns 'what does Claude Code cost' into a number you can see and steer.
Pro tips
- On a subscription, do not chase dollars: watch usage against the 5-hour and weekly limits, because that is what actually stops your work.
- On an API key, moving a routine task from Opus to Sonnet often cuts its token cost dramatically with little quality loss.
- A month-end forecast beats today's total: it tells you where the bill is heading while you can still change it.
FAQ
Common questions
Short, direct answers to the things people ask most about this.
How much does Claude Code cost per month?
On a subscription it is flat: about $20 for Pro, $100 for Max 5x, and $200 for Max 20x. On an API key it varies with usage and model; Anthropic has cited roughly $150 to $250 per developer per month on average, with power users well above that.
Why is my Claude Code bill higher than expected?
Usually because of model choice and reasoning effort on a pay-per-token key: Opus and high thinking budgets cost far more than Sonnet at low effort. Tracking per-model spend, as Tokens 4 Breakfast does, makes the driver obvious.