Claude Code Pricing 2026: Pro vs Max, Team, API, and What It Really Costs

Claude Code does not have a price. That is the first thing to get straight, because half the pages ranking for "claude code pricing" treat it like a product with its own tiers. It is not. Claude Code is bundled into every paid Claude plan, and it pulls from the same usage pool as your chats on claude.ai. So the real question is which Claude plan you buy, and whether a subscription beats paying per token on the API.

I pulled every number below from Anthropic's live pricing page and the Claude Code cost docs on 23 August 2026. Prices change, so the date matters.

Claude Code pricing at a glance

Two footnotes from the page itself: "Usage limits apply. Prices shown don't include applicable tax." and "Price and plans are subject to change at Anthropic's discretion."

One more: the Max card only says "From $100". The $200 figure for the 20x tier comes from the launch coverage in April 2025, and some regional or tax-inclusive snapshots show "From $110". Check the number at checkout, not in a blog post.

The plans, one by one

Free: chat, not Claude Code

The Free card lists chat across web, iOS, Android and desktop, code generation inside the chat UI, memory, and connectors through remote MCP. Claude Code is not on it. If you want the terminal agent, you need a paid plan or API credits.

Pro: $20/mo, and yes, it includes Claude Code

Pro is "$17 Per month with annual subscription discount ($200 billed up front). $20 if billed monthly." The card says "More usage", "Includes Claude Code", and "Includes Claude Cowork". The pricing FAQ backs it up: "Claude Code is included in all paid plans. It shares the same usage limits as the rest of your plan, so your work in the terminal and your chats draw from one pool."

Worth saying out loud because we published an article in May arguing Claude Code was leaving the Pro plan. Whatever happened in between, the live page today lists it under Pro. If you read that piece and skipped Pro on that basis, re-check.

Max: from $100/mo, 5x or 20x the usage

Max is "Everything in Pro, plus: Choose 5x or 20x more usage than Pro", "Higher output limits for all tasks", "Early access to advanced Claude features", and "Priority access at high traffic times". The FAQ adds the detail that matters for budgeting: "choose from 5x or 20x the usage of Pro per 5-hour session. Max also gives you higher output limits. Both options are billed monthly." No annual discount on Max.

Team: $20 to $125 per seat

Team is "For teams of 2 to 150". Standard seats are "$20 Per seat / month if billed annually. $25 if billed monthly." Premium seats are "5x more usage than standard seats" at "$100 Per seat / month if billed annually. $125 if billed monthly." Both seat types include Claude Code and Claude Cowork, plus central billing and admin.

If you are two co-founders, a Team Standard seat costs the same as Pro on annual billing and gets you shared admin. That is the quiet win on this page.

Enterprise: $20 per seat plus usage

"Pricing is $20 per seat per month plus usage billed at API rates, so cost scales with the models and tasks your team runs. Enterprise is billed annually." You get role-based access, SCIM, audit logs, and admins can set user and org spend limits. This is the only plan where the bill floats with usage rather than sitting at a fixed seat price.

API: pay per token, no plan limits

Claude Code also runs against a Console account with an API key. Then "Claude Code charges by API token consumption." Current base rates per million tokens (input / output) from the model pricing page: Sonnet 5 at $2 / $10, Opus 5 at $5 / $25, Fable 5 at $10 / $50, Haiku 4.5 at $1 / $5. More on what that adds up to per day below.

How the usage limits actually work

This is where people get burned, and the official wording is clearer than most third-party explanations, so here it is in full: "Every plan has usage limits that reset on a rolling five-hour session window, and paid plans add weekly limits on top. Your activity across Claude on web, desktop, mobile, and Claude Code all draws from the same pool. How much you can do depends on the length and complexity of your conversations, the model you choose, and the features you use, so there's no fixed message count."

Four practical consequences:

1. There is no message count to compare. Anyone quoting "X messages per 5 hours" for Claude Code is estimating. Anthropic does not publish one.
2. Two clocks run at once. The 5-hour window resets on its own. The weekly cap sits above it, so a heavy Monday can leave you thin on Friday even if each session stayed inside the window.
3. Chat eats your coding budget. A long research conversation in claude.ai and a long agent run in the terminal come out of the same pool.
4. Long sessions cost more than they look. Claude Code resends your whole conversation with every request. The docs list long context and cache misses as the main reasons "usage climbs in a long session", and the prompt cache lifetime is an hour on a subscription. Walk away for ninety minutes and your next message reprocesses everything.

When you hit the wall, the page gives three exits: "wait for it to reset, move to a higher plan, or, on paid plans, turn on usage credits to keep working at standard API rates." Usage credits are managed from the /usage-credits command inside Claude Code, and the cache lifetime drops to five minutes once you are drawing on them.

Anthropic also reserves the right to add "weekly and monthly caps or model and feature usage" limits "at our discretion". The comparison table lists a 200k context window on every individual plan, so the window is not what separates Pro from Max; the pool is.

Pro vs Max for coding

Same product, different pool. Here is the decision in one table.

The honest framing is that Max does not make Claude Code better. The comparison table lists the same 200k context window on Pro and both Max tiers. What $100 or $200 buys is the right to keep going. If you are not hitting limits on Pro, Max is paying for headroom you never use.

The case against Pro: the pack behind this article includes Reddit and X threads where Pro users describe hitting the 5-hour window mid-task during long agent runs. No single thread was clean enough to cite, so treat that as a pattern, not a statistic. If your work is "start an agent and let it refactor for an hour", Pro's pool will feel small faster than the price suggests.

Subscription vs API: what a day of Claude Code costs

The most useful number on Anthropic's cost page is the one almost nobody quotes: "Across enterprise deployments, the average cost is around $13 per developer per active day and $150-250 per developer per month, with costs remaining below $30 per active day for 90% of users."

That is API billing, measured across enterprise teams. Put it next to the subscriptions:

• A developer who codes with Claude 10 active days a month would spend roughly $130 on the API at that average. Pro is $20. Max 5x is $100.
• At 20 active days, the API average lands around $260. Max 20x at $200 is cheaper, and it never sends a surprise invoice.
• The 90th-percentile user (under $30 per day) at 20 days is around $600 a month on the API. That is the person Max 20x exists for.

Two caveats. First, those are enterprise averages with Opus in the mix; a solo dev on Sonnet with /clear discipline will land lower. Second, a subscription caps what you can do, the API caps only what you spend. Anthropic's own framing: "For heavy coding sessions, you can also switch to pay-as-you-go API credits through a Console account."

A practical middle path is Pro for the fixed base and usage credits for the overflow days, since credits bill at standard API rates and only when you pass the plan limit.

Ways to make either bill smaller, all from the cost docs: /clear between unrelated tasks, keep CLAUDE.md under 200 lines and move workflow instructions into skills, run Sonnet by default and reserve Opus for architecture, lower the effort level so thinking tokens (billed as output) stop eating the budget, and disable MCP servers you are not using. Background jobs cost "typically under $0.04 per session", so idle sessions are not the leak. Long context is.

Claude Code vs Cursor on price

The comparison everyone asks for, with Cursor's live pricing checked the same day:

At $20 they tie. The difference is what the $20 buys: Cursor is an IDE with an agent inside it, Claude Code is a terminal agent you point at any editor. We wrote the full Claude Code vs Cursor comparison separately; on price alone, the only clear gap is team seats, where Claude Team Standard is roughly half of Cursor Teams.

For the rest of the field, the AI coding agents directory lists current pricing for Codex, Cline, Aider and the others, and the tools directory has the cards.

Hidden costs and gotchas

• The API key trap. From the support docs: "If you have an ANTHROPIC_API_KEY environment variable set on your system, Claude Code will use this API key for authentication instead of your Claude subscription." You pay per token while your Pro plan sits unused. Unset it or use /login.
• Usage credits are opt-in but open-ended. Once on, they bill at API rates past your limit. There is a monthly spend limit in Settings > Usage; set it before you need it.
• Thinking is billed as output. Extended thinking is on by default and "the default budget can be tens of thousands of tokens per request". On the API that is real money; on a subscription it is pool you are spending.
• Agent teams multiply everything. Roughly 7x the tokens of a normal session when teammates run in plan mode.
• Tax is extra. "Prices shown don't include applicable tax."
• Refunds are the exception. "Payments are generally non-refundable, except where our Consumer Terms of Service say otherwise or where local law requires it."
• Downgrades wait for period end. Upgrades are prorated; "Downgrades take effect at the end of your current billing period."
• Cancel 24 hours early. "To avoid the next charge, cancel at least 24 hours before your renewal date."

FAQ

How much does Claude Code cost? It is not sold on its own. Pro is $20/mo ($17/mo annual, $200 up front), Max is from $100/mo, Team seats are $20 to $125 per month, Enterprise is $20 per seat plus API-rate usage. Or run it on API credits and pay per token.

What is the difference between Claude Code on Pro and on Max? Nothing in the product. Max gives you 5x or 20x the usage of Pro per 5-hour session, higher output limits, early feature access, and priority at peak times. Pro is $20/mo, Max is from $100/mo and monthly only.

Is there a free plan for Claude Code? No. Free covers chat and in-chat code generation. The FAQ says Claude Code is "included in all paid plans", and the Free card does not list it.

Does Claude Code pricing include API usage? A plan includes usage up to its limits. Past that, usage credits bill at standard API rates. An ANTHROPIC_API_KEY in your environment bypasses the subscription entirely and bills the API.

Can I get a refund? Generally no. The page says payments are non-refundable except where the Consumer Terms or local law say otherwise, with an in-app "Claude Refund Request" flow to check eligibility.

Is Claude Code cheaper than Cursor? Same price at the entry tier, $20 each. Claude Team Standard ($20 to $25) is cheaper than Cursor Teams ($40). Max at $100 or $200 buys usage, not a different tool.

What are the Claude Code rate limits on Pro? No published count. A rolling 5-hour window, a weekly cap on paid plans, and "at least 5x more usage per 5-hour session than Free". Usage depends on conversation length, model, and features.

Does Claude Code offer annual billing? Pro yes, $17/mo billed $200 up front. Max no, "Both options are billed monthly." Team has both, Enterprise is annual.

Is Claude Code worth the subscription for daily coding? For a daily user, yes at the Pro level: Anthropic's own API average of about $13 per developer per active day means ten active days already costs more than Pro. Step up to Max only when the 5-hour window is what stops you, not the model.

How do I upgrade or cancel? Settings > Billing on web or desktop. Upgrades are prorated, downgrades apply at period end, and cancellation keeps the plan active to the end of the cycle. App Store and Google Play subscriptions are managed in those apps.

Is Claude Code worth it?

If you already pay for Claude Pro, the agent is free to you and the only question is whether the pool is big enough. Try it for a month before touching Max. Most of the "Pro is too small" stories are long sessions without /clear, and that is fixable for $0.

If you are choosing between a $20 plan and the API, the plan wins for anyone who codes with it more than a couple of days a month. The API wins for bursty users and for anyone who needs to run past a weekly cap without asking permission.

Where this recommendation is wrong: if you run several agents in parallel, or you work in bursts of sixty-hour weeks followed by nothing, a subscription's weekly cap is the wrong shape for you and pay-as-you-go is the better call even when it costs more. And if you live inside an IDE and want the agent in the sidebar, Cursor at the same $20 is the closer fit, price tie or not.

Want the current card with links and alternatives? See Claude Code in the tools directory, then the AI coding agents directory for what else $20 buys.