Which Claude plan you need for 2026?

Wich Claude plan you need?

Plenty people wey dey ask which Claude plan dem need dey stuck for the same small set of decisions. Claude Code dey inside Pro plan, or na separate purchase? Developer suppose pay monthly or per token? Team of five people suppose buy five Pro accounts or Team plan? I don answer different versions of these questions for support tickets for one year now, mostly from customers wey dey set up Claude Code for one of our servers. So, na this decision guide I wish I fit just link.

The lineup as of July 2026 be: Free, Pro, two Max tiers, Team (with two seat types), and Enterprise, plus the API. API no be plan at all. Na separate pay-per-token account e be. Most individual users no need am, but most developers go eventually use am. The question people ask pass get one-line answer: Claude Code dey included for every paid plan, starting from Pro, and e no dey inside free tier. Everything after that na about matching the way you use am with the correct tier.

Prices dey change. I go quote the numbers wey dey for claude.com/pricing the day I write this, and I go include dates. Treat that page as the source of truth, no be this post and no be the many third-party posts wey don already become outdated.

Wetin really dey control the tiers

Remove the marketing talk, four things dey separate the plans:

• Usage limits. Na this be the real product. Every tier above Free dey priced as a multiple of Pro allowance. Max dey sold literally as "5x" and "20x" Pro usage for each session. Under the hood, these limits na token accounting. Na why heavy Claude Code session dey use dem finish faster than chat ever go, how Claude Code dey use tokens dey mostly come from context wey e resend, no be from wetin you type.
• Claude Code inclusion. Paid plans get am, Free no get am. Pro, Max, Team (both seat types), and Enterprise all cover Claude Code for terminal and supported IDEs, VS Code, Cursor and other forks, JetBrains, with the same shared limit as chat. Which of these surfaces dey available at all depend on your operating system, no be your tier, and Linux get the CLI and editor plugins complete, while desktop app still dey beta wey you install from apt. Something wey plan include no mean say e free. So if you want the monthly amount for your own usage instead of simple yes or no, wetin Claude Code really cost for each plan go calculate am with sample month.
• Feature and model access. Pro and higher tiers get the complete model picker and unlimited projects. Max add higher output limits and priority access to new models when traffic high. Free tier no be useless reduced version, though. As of July 2026, e include web search, file creation, code execution, connectors, and extended thinking.
• Admin surface. SSO, central billing, and admin controls start for Team. SCIM, audit logs, the Compliance API, custom data retention, and customer-managed encryption keys start for Enterprise. If these acronyms no mean anything to you, you no be Enterprise customer, and e dey okay.

Notice wetin no dey this list: intelligence. Free user and Max user mostly dey talk to the same models. You dey buy more usage and supporting infrastructure, no be smarter Claude.

Casual user: stay for Free until e begin hurt

If you dey use Claude few times every week for drafting, explaining, and occasional script, Free genuinely enough, and I go tell you make you stay with am. The limits tight, but dem dey reset, and the features cover most conversational use.

Two things suppose push you go Pro. The first na hitting the usage wall for middle of task more than once every week; if dem cut you off for middle of something wey you dey pay attention to, $20 worth am to stop that. The second na if you want Claude Code at all, because Free simply no get am. As of July 2026, Pro dey cost $20/month when dem bill you monthly, or $17/month with annual billing. The annual discount na real money if you know say you go stay, but I go first use one monthly cycle to confirm say the limits fit you. Trying a tier low-risk for both directions anyway, because going back to cheaper plan or cancelling completely still leave you with the month wey you don already pay for. If you want the full inventory before you commit, wetin Pro add on top free plan and where its limits go stop you na the longer version of this paragraph. And if your real question na which $20 subscription to buy, instead of whether to buy one at all, Claude and ChatGPT ladders dey align closely enough tier by tier say the answer usually depend on how much coding you dey do.

Wetin Pro no fix: e no go make limits disappear. Pro allowance dey shared between Claude chat and Claude Code, so the day you start to use Claude Code seriously na the day Pro go start feel small. That no be defect; na the tier boundary dey do the work wey dem design am for.

Developer: subscription or API?

This na fork wey dey cause pass confusion. Both options dey run the same models, but their pricing structure dey opposite. So whether API cheaper pass subscription depend on how much you dey use am.

Subscription na fixed fee with limit. You pay $20 (Pro) or $100/$200 (Max 5x / Max 20x, as of July 2026). You get usage allowance wey dey reset on rolling basis. Some plans still get weekly caps on top. The support docs explain the current windows. When you reach the limit, work go stop until reset, unless you clearly agree to pay for extra usage credits. Anthropic make am clear say per-token billing no go start automatically without your consent. If e stop you in the middle of task, find out which window you dey wait for first. Na that one go decide whether you switch to smaller model, reduce the context, buy credits, or wait for reset.

API na meter without limit. You pay per million tokens, for input and output. As of July 2026: Haiku 4.5 na $1 input / $5 output; Sonnet 5 na introductory $2 / $10 until August 31, 2026, then $3 / $15; Opus 4.8 na $5 / $25; and Fable 5, the premium tier, na $10 / $50. Batch processing cut all these prices by half. Prompt-cache reads cost roughly one-tenth of the input price. Fable 5 na the line item wey dey make people worry, so know which jobs really justify the rate before you send steady traffic go am. If you hope say you fit test am free first, API no get free tier. Na only small signup credit and some free components dey available. Plan to pay from your first real script. Check claude.com/pricing before you prepare budget. The Sonnet introductory price alone fit change before many of una read this.

The decision rule wey I give customers dey depend on who starts the request:

1. Human for keyboard, wey dey work interactively, you, inside Claude Code, throughout the day. Use subscription. Interactive coding na exactly the bursty, high-volume, unpredictable usage wey flat pricing dey designed to handle. One serious day of agentic coding fit resend tens of millions of context tokens. With API rates, the invoice fit scary. With Max, na normal Tuesday. Start with Pro. If you dey reach the limit most days, use Max 5x. If you dey run parallel sessions or long autonomous tasks, use Max 20x. If you dey between those two, the $100 and $200 tiers differ only for how much usage you get, not for models or features. This one apply whether Claude Code dey run for your laptop or server. Plenty people dey keep Claude Code running on a VPS inside tmux so long sessions fit continue after SSH connection drop. Billing still dey use the same subscription either way.
2. Program wey dey start requests by itself, like cron job, webhook handler, chatbot, or anything server-side. Use API, no exception. Subscriptions na consumer accounts with consumer terms. Automation belong on API key with per-token billing wey fit scale from small experiments to production traffic. If na this you wan do, start with your first Claude API app on a VPS, including the account, key, and first working script.
3. Both, na the common final setup. Use Max plan for your own daily coding and API key for the things you build. Dem be separate accounts with separate billing, and this separation na advantage.

One failure mode dey catch people regularly, so make we state am exactly: if ANTHROPIC_API_KEY dey set for your environment, Claude Code go authenticate with that key and charge your API account with per-token rates. E go silently ignore your Pro or Max subscription. You go discover am when the Console invoice arrive. Before you assume say Claude Code dey use your subscription:

echo "${ANTHROPIC_API_KEY:+key is set - Claude Code will bill the API}"
unset ANTHROPIC_API_KEY   # and remove it from ~/.bashrc or ~/.profile

Inside Claude Code, /status go show the account wey you authenticate against. Check am once for every new machine. For VPS especially, old key wey somebody export inside dotfile from previous project na the usual cause.

Team: five people na the threshold

Team dey for groups of 5 to 150 people. As of July 2026, e get two seat types: standard seat wey cost $20/month annual ($25 monthly) and get about 1.25x Pro per-session usage, plus premium seat wey cost $100/month annual ($125 monthly) and get 6.25x. Na premium seat dem design for people wey dey work inside Claude Code, and both seat types include am. You fit mix the seats, and na this be the main benefit: two developers for premium, three writers for standard, with one invoice.

The comparison with individual plans no get any special difference, and na intentional. Standard seat cost about the same as Pro. Five personal Pro accounts and five standard seats go reach roughly the same monthly amount, so price no be the thing wey decide this choice. Wetin you dey really buy na the management setup: central billing instead of five expense reports, SSO, admin control over who get access, and default policy wey no allow training with your data. Individual consumer accounts no contractually promise this to your employer. Five personal accounts no give you any of these things, and nobody fit cut off person wey leave the company on the person last day. Once person ask, “company fit pay for this?”, Team na the answer. The other option na shared logins and person own card, and both options dey end badly.

Pay attention to whether the seat type fit the user. Developer wey dey reach standard seat weekly cap every week na premium user wey get wrong badge; upgrade go cheaper than afternoons wey work stop. Premium seats get two weekly limits: one for all models together, and one specifically for Sonnet-class models. Heavy users should quickly check the Team plan docs before dem assume say the multiplier na the whole matter. If several people dey reach those weekly caps every week, buying more seats no go solve the problem. At that point, the question don turn to Enterprise.

Enterprise: when Team no longer enough

Enterprise don change shape, and plenty third-party pricing posts never update. As of July 2026, e no be "contact sales for a mystery number": self-serve tier dey with 20-seat minimum wey you fit buy directly, plus sales-assisted tier for bigger contracts. The pricing model dey opposite to consumer plans: small per-seat fee (about $20/seat/month, billed annually) wey cover access only, while all usage dey billed separately at API rates. No per-seat usage ceiling at all. Your heaviest Claude Code user no go hit any limit, and your bill go follow actual consumption instead of guessing based on seat count. To budget this kind arrangement, forecast token consumption instead of counting people. So, before you carry any figure go finance, e good make you read wetin the $20 seat really cover and wetin the meter charge on top.

Na this reversal show who suppose buy am. If your organization get problem with predictability, where one fixed line item dey apply to each person, Team na your plan. If the problem na capacity, power users dey finish premium-seat limits, or compliance like audit logs, SCIM provisioning, Compliance API, custom data retention, customer-managed encryption keys, US-only inference, or HIPAA-ready configuration, Enterprise na the only tier wey get any of dem. Nobody dey buy Enterprise because of the chat experience; dem dey buy am because security review reject every other option. And if that review dey compare vendors instead of tiers, na only one of the two publish Enterprise seat price at all, so seat minimums, retention terms, and metered agent usage na the things wey really worth put side by side.

Di whole decision na six lines

• Occasional chat, no coding: Free, until the limits begin annoy you.
• Regular use, or you want Claude Code at all: Pro.
• Claude Code for most of the day, and you dey hit Pro limit: Max 5x, then 20x.
• Anything programmatic, bots, pipelines, server-side calls: API, no matter which subscription you also get.
• 5+ people, company money, admin needs: Team, wey mix standard and premium seats.
• Compliance requirements or power users wey no want usage limit: Enterprise.

And whichever branch you choose, verify the two numbers wey really matter for claude.com/pricing on the day you buy: the price of your tier and how its usage limit dey work. Both don change for the last twelve months. Both go change again.

FAQ

Claude Code dey inside Pro plan?

Yes. As of July 2026, the $20/month Pro plan include Claude Code for terminal and supported IDEs (VS Code and im forks, JetBrains). Dem dey charge am against the same usage allowance as Claude chat, so no separate Claude Code subscription dey. But if exported ANTHROPIC_API_KEY environment variable dey set, Claude Code go bill your API account instead. Unset am if you want make your subscription cover the usage.

Free tier dey include Claude Code?

No. As of July 2026, the free plan cover Claude for web, desktop, and mobile, including web search, file creation, and code execution, but e no cover Claude Code. To use Claude Code, you need either paid plan (Pro and above) or API key with pay-per-token billing. For regular use, Pro na the cheapest way to start.

I suppose get Claude subscription or use the API?

Decide based on who dey initiate the requests. Human wey dey work interactively, especially for Claude Code, suppose use subscription because flat pricing fit absorb the huge token volume wey agentic coding dey use. Program wey dey make requests by itself, like cron jobs, chatbots, and pipelines, suppose use API. For there, per-token billing dey scale with load, and no consumer usage ceiling dey. Most developers eventually get both, and na so the setup suppose be.

Wetin Claude Max give pass Pro?

Na volume and plumbing, no be smarter models. Max dey give multiples of Pro usage allowance (5x or 20x per session as of July 2026), plus higher output limits and priority access when network traffic plenty. Which Max tier you need depend on how often Pro limit dey stop you: choose 5x if na most days, and 20x if you dey run parallel sessions or long autonomous tasks.

Team or Enterprise: which one my company need?

Decide based on wetin dey break. If you want predictable cost per person, with SSO and one invoice for group of 5 to 150, Team na the plan. You fit mix standard and premium seats inside am. If your heaviest users dey always reach the weekly caps, or security review dey request SCIM, audit logs, custom data retention, or HIPAA-ready configuration, Enterprise na the tier wey fit handle am. As of July 2026, Enterprise billing still different: modest per-seat fee cover access, while usage dey billed separately at API rates with no per-seat ceiling.