Claude Code Cost: Plans, API Prices and Real Math

Claude code cost starts at $20 a month. As of October 2026, Claude Code is included in every paid Claude plan: Pro ($20 monthly or $17 a month billed yearly), Max 5x ($100), Max 20x ($200), Team seats from $20 and Enterprise. The free Claude plan does not include it. You can also skip the subscription and pay per token through the API, where a typical day of heavy agent work can cost more than a whole month of Pro.

The short version: Pro fits a few focused sessions a week, Max 5x fits daily coding, and the API makes sense for automation and for teams that want billing by usage. The math below shows why.

How we researched this: official pricing pages, Anthropic's docs and help center articles as of October 2026, plus public user reports. We haven't run our own benchmarks, so the cost examples below are calculations from list prices, not measured bills.

How much does Claude Code cost per plan?

Here is every way to pay for Claude Code, from Anthropic's pricing page and the Max plan help article, checked October 24, 2026.

A few details matter more than the sticker price:

• Limits are shared. Anthropic's Pro and Max help article says the 5-hour session limit and the weekly limit are shared between Claude chat and Claude Code. A long chat session eats into your coding time.
• There are two windows. Usage resets every five hours, and paid plans also have a weekly cap on top.
• Max includes API credits, but not for Claude Code. Max 5x gets $100 and Max 20x gets $200 of monthly API credits. Anthropic's API credits article says they cover the API, the Agent SDK and claude -p run with an API key, but not interactive Claude Code in the terminal, IDE, desktop app or web. Unused credits expire each billing cycle.
• Going past the limit costs API rates. On paid plans you can turn on usage credits to keep working after you hit a limit, billed at standard API prices.

Anthropic doesn't publish the plan limits in tokens or messages. That's the main reason "is Claude Code expensive" has no single answer: you can't convert Pro into an exact number of tasks. For what every plan includes beyond coding, see our Claude pricing guide.

What does Claude Code cost on the API?

If you sign in with a Claude Console account instead of a subscription, Claude Code bills per token. These are the list prices per million tokens for the current models, from claude.com/pricing as of October 2026:

Two things drive the bill in Claude Code. First, it sends your whole conversation with every request, so long sessions get expensive even when your next question is short. Second, prompt caching makes that resent history cheap: a cache read on Sonnet 5.5 costs $0.10 per million tokens, 20 times less than fresh input.

Anthropic's own figure is a useful sanity check. The Claude Code cost docs say the average across enterprise deployments is about $13 per developer per active day, or $150 to $250 a month. 90% of users stay under $30 per active day.

Worked examples: API vs subscription

To compare plans, we need a unit. We'll use one "task": a medium bug fix or small feature that takes Claude Code about 30 model requests (reading files, editing, running tests, fixing). These are our assumptions, not Anthropic numbers:

• Each request carries about 55,000 tokens of cached context (read from cache).
• Each request adds about 5,000 new tokens that get written to the cache.
• Each request produces about 1,500 output tokens, including thinking. Anthropic's docs note that thinking tokens bill as output, and current models always use extended thinking.

Here is the math for one task at October 2026 list prices.

Sonnet 5.5, one task:

• Cache reads: 30 × 55,000 = 1.65M tokens × $0.10 = $0.165.
• Cache writes: 30 × 5,000 = 150,000 tokens × $2.50 = $0.375.
• Output: 30 × 1,500 = 45,000 tokens × $10 = $0.45.
• Total: about $0.99.

Opus 5.5, one task: every Opus 5.5 rate is exactly double Sonnet 5.5, so the same task costs about $1.98.

Your real numbers will move with codebase size, how often you /clear, and how much the model thinks. But the shape holds: output and cache writes cost more than the big pile of cached reads.

Now scale that to a month:

The daily-developer row lands close to Anthropic's own $150 to $250 monthly average, which suggests the assumptions are in the right range.

The takeaway: if your usage fits inside a plan's limits, a subscription is much cheaper than the API. The catch is the "if". We can't tell you whether 200 tasks fit inside Max 5x, because Anthropic doesn't publish the limits in tokens. Its /usage command shows how close you are to your limits, so the practical test is to start on a plan and watch it for a week.

Which Claude Code plan fits which kind of user?

Pick by how often you hit limits, not by features. Every paid plan gets the same Claude Code tool.

• Pro ($20) fits students, hobbyists and developers who use Claude Code a few times a week. At API rates, about 20 of our Sonnet-sized tasks would already cost $20, so Pro pays for itself fast if the limits hold up for you.
• Max 5x ($100) fits anyone using Claude Code most workdays. If Pro's 5-hour limit stops you more than once or twice a week, this is the next step.
• Max 20x ($200) fits people running long agent sessions, several sessions in parallel, or subagents and agent teams. Anthropic's docs say agent teams use roughly 7x more tokens than a normal session when teammates run in plan mode.
• Team fits companies that want central billing and admin controls. Standard seats cost about the same as Pro; Premium seats cost about the same as Max 5x.
• The API fits CI pipelines, scripts using claude -p, and organizations that want spend limits per workspace instead of a fixed seat. It's also the fallback when you need more than Max 20x allows.

If you're also weighing editors, our Cursor vs Claude Code comparison looks at what $20 buys on each side, and the Cursor pricing guide breaks down Cursor's new plans.

Is the $20 Claude Code plan worth it?

Yes, for most people starting out. Pro gets you the full Claude Code tool, the Opus, Sonnet and Haiku models, and enough usage for regular part-time coding. Our worked example puts 20 medium Sonnet tasks at about $20 on the API, and Pro also covers your Claude chat use.

The downside is the ceiling. Pro's limits are shared with chat, there's a weekly cap on top of the 5-hour window, and Anthropic's newest Fable model is listed as usage credits only on Pro. If you code with it all day, you will hit the limit and either wait, pay API rates through usage credits, or move to Max.

Is Claude Code expensive?

It depends on how you pay. On a subscription, Claude Code costs the same as a normal Claude plan, $20 to $200 a month. On the API, it can be expensive: Anthropic's own enterprise average is about $13 per active developer day, and our heavy-user example passes $650 a month on Sonnet 5.5.

Compared with other tools, the $20 entry price matches most AI coding assistants. What's different is how fast heavy agent use climbs. For other options, see our roundup of the best AI for coding and how Claude Code compares with Codex.

How to keep your Claude Code bill down

Most of these come from Anthropic's cost management docs:

1. Use /clear between unrelated tasks. Old context gets resent with every request. Clearing costs nothing.
2. Default to Sonnet. Anthropic says Sonnet handles most coding tasks well. Switch to Opus with /model for hard design problems only. In our example, that halves the cost.
3. Lower the effort level for simple work. Thinking bills as output, the most expensive token type. Use /effort to turn it down.
4. Trim CLAUDE.md. It loads into every session. Anthropic suggests keeping it under 200 lines and moving workflow-specific instructions into skills.
5. Check /usage. On Pro and Max it shows what's eating your limits, including skills, subagents and MCP servers, with suggested fixes.
6. Unset ANTHROPIC_API_KEY if you're on a plan. Anthropic warns that when this variable is set, Claude Code uses the API key and you pay API rates instead of using your subscription.

FAQ

How much does Claude Code cost?

As of October 2026, Claude Code comes with Claude Pro ($20 a month, or $17 billed yearly), Max ($100 or $200 a month), Team seats from $20 and Enterprise. On the API it's pay per token, for example $2 input and $10 output per million tokens on Sonnet 5.5.

Is Claude Code free?

No. The Claude Free plan doesn't include Claude Code. The cheapest way in is Pro, or a small amount of API credit if you only want to try it. Our guide to whether Claude Code is free covers every cheap option, including local models.

What's the Claude Code price for teams?

Team Standard seats cost $25 a month or $20 a month billed yearly. Premium seats cost $125 or $100 and get 5x the Standard usage. Enterprise is $20 per seat a month with a 20-seat minimum, plus usage billed at API rates.

Do Max plan API credits pay for Claude Code?

Not for interactive use. The $100 or $200 of monthly credits cover the API, the Agent SDK and claude -p with an API key. Interactive Claude Code in the terminal, IDE, desktop or web draws on your plan limits instead.