Claude Code Pricing 2026: Every Plan, API Rates and Real Monthly Cost

Claude Code comes with every paid Claude plan, not the Free one. Pro costs $20 a month, or $17 a month billed annually ($200 up front), Max costs $100 or $200 a month, Team seats run $20 to $125, and the API bills per token (Sonnet 5.5: $2 input, $10 output per million). Verified on claude.com in October 2026.

My take: for one person coding a couple of hours a day, Pro is the cheapest way in and Max 5x is the realistic plan once Claude Code becomes your main editor. The API only wins when your usage is light and irregular, or when a company wants spend caps per developer instead of seats.

Claude Code plans and prices at a glance

Prices exclude tax. Every figure above was read on Anthropic's pricing page and help center on 28 October 2026.

Dupple data: Toolradar, our software directory, tracks 181 AI coding tools as of October 2026, and the median cheapest paid plan among the 59 that publish one is $19 a month. Claude Pro sits right on that median, so you are not paying a premium to get Claude Code.

What changed in Claude Code pricing this year

Three things moved in 2026, and most pricing explainers online still quote the old numbers.

First, the model line got cheaper per token. Opus 5.5 lists at $4 per million input tokens, against $5 for Opus 4.8, which Anthropic now lists as a legacy model, and its output rate fell by a fifth too. Sonnet 5.5 dropped to $2 and $10 from Sonnet 4.6's $3 and $15. If you read a guide that prices Claude Code API use at $3/$15, it is pricing last year's model.

Second, both Max tiers now ship with monthly API credits worth the same as their subscription price. That reads as if Max hands back part of your Claude Code bill, but it does not, and the section on Max below explains why.

Third, Claude Code is in every paid plan, from Pro up. There is no separate Claude Code subscription to buy, and no "Claude Code Pro" tier, whatever some comparison sites imply.

Is Claude Code included in Claude Pro?

Yes. Pro includes Claude Code in the terminal, the IDE extensions, the desktop app and the web. The catch is that Claude Code and your Claude chats draw from one usage pool, in Anthropic's own wording on the pricing page FAQ. If you spend the morning drafting a report in Claude chat, you have less left for coding in the afternoon.

That shared pool is the single most important fact on this page for people who use Claude for more than code. A product manager who writes specs in chat and then asks Claude Code to build a prototype is spending one allowance twice, while a developer who only codes in the terminal barely notices the pool.

Pro gives at least five times the per-session usage of the Free plan, and paid plans add weekly limits on top of the five-hour rolling window. Anthropic publishes no message or token count for any plan, only these multiples. "No fixed message count" is the exact phrase on the pricing page, because usage depends on context length, the model and the features you use.

If you want to try before paying, how to get Claude Pro for free lists the legitimate routes.

Claude Code Pro vs Max: which plan do I need?

Pick Max when the five-hour limit interrupts real work more than once or twice a week. Max 5x gives five times Pro's per-session allowance, and Max 20x gives twenty times it for double the price. Both are billed monthly only, so there is no annual discount to chase.

The feature gap between the two is small, so you are mostly paying for hours. Max adds higher output limits, priority access at busy times, early access to new features and Fable (Anthropic's top model) at up to 50% of the weekly limit. On Pro, Fable runs only through usage credits billed at API rates. For most coding, Opus 5.5 and Sonnet 5.5 do the work, so Fable alone is rarely the reason to upgrade.

We cover the plan choice for general Claude use in Claude Pro vs Max. For Claude Code specifically, the deciding variable is how many hours a day the agent runs, not how many features you need. If you run several Claude Code sessions in parallel, Max 20x is the plan built for that pattern.

The Max API credits do not cover Claude Code

The monthly credits on Max work on the Claude Platform: the Messages API, the Console Playground, Managed Agents and the Agent SDK. Anthropic's monthly API credits article states they do not apply to interactive Claude Code in the terminal, IDE, desktop or web, and they do not raise your plan's usage limits.

They also expire at the end of each billing cycle. So the credits are worth real money only if you build something against the API, such as an internal agent or a script that calls Claude. If you only use Claude Code, treat Max as a pure usage plan and ignore the credits when you compare costs.

How much does Claude Code cost with the API?

The API route bills every token Claude Code sends and receives at list price, with no session limit. You sign in with a Claude Console account instead of your subscription. Here are the per-million-token rates for the current models:

Source: Claude pricing, API tab, read 28 October 2026. US-only inference costs 1.1x these rates, and fast mode for Opus 5.5 costs 2x.

Cache reads matter more than the headline input price. Claude Code rereads your project context on almost every turn, and most of that comes from the prompt cache at a twentieth of the input price. That is why real bills are lower than a naive "tokens times input price" estimate. Model choice is the other big lever: Claude Opus vs Sonnet covers when the dearer model earns its rate.

Anthropic also publishes what companies actually spend. Its Claude Code cost documentation says enterprise deployments average around $13 per developer per active day and $150 to $250 per developer per month, with 90% of users staying under $30 per active day.

Real monthly cost: subscription vs API for one developer

Here is the scenario: one person codes with Claude Code 2 hours a day, 20 working days a month. To price the API side, we assume a day of that work uses 20 million cache-read tokens, 1 million cache-write tokens and 300,000 output tokens. That is an assumption, not a measurement, chosen so the daily cost lands near Anthropic's own published average.

Our arithmetic at list prices:

• Sonnet 5.5: 20 x $0.10 + 1 x $2.50 + 0.3 x $10 = $7.50 a day, so $150 a month and $1,800 a year.
• Opus 5.5: 20 x $0.20 + 1 x $5 + 0.3 x $20 = $15 a day, so $300 a month and $3,600 a year.

The break-even is stark. At this volume, the API on Sonnet 5.5 costs one and a half times Max 5x, and on Opus 5.5 it costs three times as much. A subscription wins for anyone who codes daily, as long as the limits hold.

The API wins in two cases. If you code two or three days a month, a light month on Sonnet 5.5 can come in under the Pro price. And if you need a hard ceiling per person, the API lets an admin cap spend, which no individual plan does.

How much does Claude Code cost for a team?

A five-person team on Team Standard seats pays 5 x $20 x 12 = $1,200 a year billed annually, or 5 x $25 x 12 = $1,500 billed monthly (our arithmetic). The annual commitment saves $300.

Premium seats give five times a Standard seat's usage. Five Premium seats come to $6,000 a year annually, or $7,500 monthly. Most teams mix seat types: three Standard and two Premium seats cost $260 a month, or $3,120 a year on annual billing.

Team also pools monthly API credits equal to each seat's annual-billing price, capped at $500 a month. That same three-plus-two team gets $260 a month for API work, though, as on Max, the credits do not cover interactive Claude Code.

Enterprise flips the model from flat seats to metered usage. You pay the same per-seat fee as a Team Standard seat, on an annual contract with a 20-seat minimum, so at least $4,800 a year in seat fees, plus every token at API rates. With Anthropic's published average of $150 to $250 per developer per month, usage would add $3,000 to $5,000 a month for 20 developers (our arithmetic). Enterprise buys admin controls and spend limits, not a discount.

What happens when I hit the Claude Code limit?

Claude Code stops and shows when your session resets. You then have three options: wait for the five-hour window to roll over, move to a higher plan, or turn on usage credits to keep working at standard API rates. You can see where you stand at any time in Settings > Usage.

Usage credits are the hidden cost of Pro. They are useful for the occasional crunch day, but a Pro subscriber who turns them on every week may end up paying more than Max 5x without noticing. Check your monthly credit spend: once it plus your Pro fee nears the Max 5x price, Max 5x is the cheaper plan.

Anthropic also reserves the right to add weekly or monthly caps and to limit specific models or features "at our discretion". That clause is on the pricing page, and it means no plan guarantees a fixed amount of Claude Code time. The Claude Code cost docs list the ways to stretch an allowance, such as clearing context between tasks and using Sonnet for routine edits.

Hidden costs and weaknesses

• One pool for chat and code. Heavy chat use on the same account eats your Claude Code time, on every plan.
• No published limits. Anthropic gives multiples (5x, 20x) but no token or hour figure, so you cannot compare Pro's allowance with a competitor's in numbers.
• Max is monthly only. There is no annual discount on Max 5x or Max 20x, unlike Pro and Team.
• Credits that look like a discount are not one. The Max and Team API credits never pay for interactive Claude Code.
• Enterprise has a floor. The 20-seat minimum puts the entry point at $4,800 a year before usage.
• Taxes are extra. Every listed price excludes applicable tax.

Claude Code is still a strong buy at these prices; the agent is why many developers pay for Claude at all. For how it stacks up against rivals at similar prices, see Claude Code vs Cursor and Codex vs Claude Code. The broader field sits in our best AI coding assistants guide, and the best Claude Code skills and best MCP servers for Claude Code show what to add once you are set up. If ChatGPT is your other option, Claude vs ChatGPT for coding compares the two. Developers who want a terminal-first take on the same numbers can read Devshot's Claude Code pricing breakdown. Toolradar keeps a Claude Code profile with alternatives.

Methodology

Every price on this page was read on claude.com/pricing on 28 October 2026, including the Individual, Team and Enterprise, and API tabs. The API credit rules come from Anthropic's help center, and the average spend figures from Anthropic's Claude Code documentation. The developer scenario uses an assumed token volume, stated above, priced at list rates; totals are our arithmetic. The Toolradar count comes from its live AI coding category. We recheck prices every 60 days.

Cite this: Dupple, "Claude Code Pricing 2026: Every Plan, API Rates and Real Monthly Cost", November 2026.

FAQ

Start on Pro, track how often Claude Code hits the limit for two weeks, and move to Max 5x only when waiting costs you real working time. Start the Dupple X trial →