Claude Code Pricing 2026: Every Plan, the Free Tier and the Catch

Claude Code pricing in 2026 runs from $0 for chat-only access up to $200 a month on the Max 20x plan, with Team seats priced between $20 and $125 per person and Enterprise billed per seat plus metered API usage. Most solo developers settle between the $20 Pro plan and the $100 Max 5x plan.

The catch, and the reason this keeps tripping people up: the free plan does not include Claude Code at all. “Free” only ever applied to Claude’s chat product. If you want the coding agent, you are paying for it in one form or another, whether that is a flat subscription seat or metered API tokens.

This guide breaks down every current plan, what each one actually includes, how the token-based API alternative prices out, and where teams tend to get surprised by their bill. Figures below are current as of September 2026 and sourced from Anthropic’s official plan chooser and Claude’s pricing page. This page is scoped to Claude Code specifically; for the plans compared across Claude’s whole product line, including features Claude Code does not touch, see our full Claude pricing guide.

How Much Does Claude Code Cost Per Month?

Claude Code costs $0 to run only inside a Pro plan or higher; there is no standalone price because it is bundled into a subscription seat or billed per API token. In practice, expect $20 to $100 a month for individual use, $20 to $125 per seat for teams, and roughly $150 to $250 per developer per month for heavy Enterprise API usage.

Anthropic’s own enterprise telemetry, cited in Claude Code’s cost management documentation, puts the average organizational cost at about $13 per developer per active day, with 90% of users staying under $30 on any given day. That is a wide range because a quick bug fix and an eight-hour refactor draw on the same meter completely differently.

Three variables decide where you land on that range: which plan or billing model you choose, which Claude model handles your requests (Haiku is roughly a fifth the cost of Sonnet, which is itself a fifth the cost of Opus), and how large your working context is, since Claude Code resends conversation history with every tool call.

What Are the Claude Code Pricing Plans, From Free to Enterprise?

Claude Code pricing splits into two families: flat-rate subscription seats (Free, Pro, Max, Team, Enterprise) and metered API billing charged per token. Subscriptions cap what you pay but ration usage in five-hour and weekly windows; the API has no ceiling but scales linearly with every request you send.

The table below reflects Anthropic‘s published rates as of September 2026.

Team plans dropped their seat minimum to just two seats, according to an analysis of the current Team plan structure, down from the five-seat minimum Anthropic originally required. That means a two-person shop can now get Team-level admin controls, SSO and a “no training on your content” data policy for $40 a month on annual billing, which used to require a much bigger commitment.

Is There a Free Version of Claude Code?

No. Claude has a free chat plan, but it explicitly excludes the Claude Code CLI, SDK and IDE integrations. The cheapest way to actually run Claude Code is the $20-a-month Pro plan; anything billed as “free Claude Code” is either a guest pass, a time-limited trial, or a workaround using someone else’s API credits.

This is the single most common point of confusion in claude code pricing searches. Chat and code are different products with different access tiers, and Anthropic keeps that line firm: the free plan gets you web chat, file creation, code execution in a sandbox, and artifacts, but not the terminal-based coding agent.

What Exactly Is Included in the Claude Code Free Tier?

Strictly speaking, there is no Claude Code free tier; the phrase describes what the Free chat plan includes instead, which is unlimited-feeling but rate-limited access to Claude in a browser, desktop or mobile app, plus web search, memory and basic file handling. None of that reaches the command line.

Where people get a taste of Claude Code without paying is through two narrow official paths. A current Max subscriber can send an eligible contact a seven-day guest pass carrying Pro-level access, including Claude Code, though the recipient has to be new to paid Claude plans and hand over payment details before the trial starts. Separately, new Claude Console (API) sign-ups receive a small batch of free credits meant for verifying a setup works, not for real project work, as detailed in a review of Claude Code’s official trial options from April 2026.

Is There a Claude Code Free Trial?

Not a public, always-on one. Anthropic offers no self-serve “free for 14 days” button for Claude Code the way many SaaS products do. The realistic options are the Max-subscriber guest pass mentioned above, a handful of throwaway API test credits, or simply starting on the $20 Pro plan, which you can cancel before the next billing cycle if it is not a fit.

Given how quickly a genuine coding session burns through a small credit balance, budget for the Pro plan as your evaluation tier rather than hunting for a workaround. Twenty dollars buys a real week of usage; a trial credit balance typically does not survive a single afternoon of active refactoring.

Does Claude Code Charge Per User or Per Usage?

Both, depending on how you buy it. Subscription plans (Pro, Max, Team, Enterprise seats) charge a flat price per person per month and ration usage in five-hour rolling windows plus a weekly cap. The API and Claude Console charge purely per token consumed, with no per-seat fee and no usage ceiling.

Every subscription tier shares its usage allowance across Claude Code, Claude.ai chat and Cowork, so a developer who also uses Claude in the browser for planning docs draws from the same pool their coding sessions use. Teams that mix the two access methods, some members on seats and some authenticating with a raw API key, get billed differently per person, which is worth flagging to finance before rollout.

What the Underlying API Actually Costs Per Model

If you connect Claude Code to a Console (API) account instead of a subscription, you pay Anthropic’s list price per million tokens, and the model you pick changes the bill dramatically.

Rates and figures above are drawn from Anthropic’s published API pricing. Prompt caching cuts the reprocessing cost of unchanged context by up to 90%, and batch API requests get a flat 50% discount on both input and output tokens, which matters if you run scheduled or overnight agent jobs rather than live interactive sessions.

Opus 5.5 succeeded the two-month-old Opus 5 on September 22, 2026, at a lower price than the model it replaced ($20 rather than $25 per million output tokens), and Anthropic says it matches the pricier Fable 5.1 on most coding work. That claim held up in early developer testing: one engineer documented in Every’s vibe-check roundup switched Opus 5.5 in as his daily driver in place of Fable 5.1, estimating it delivers roughly 90 percent of Fable’s coding capability at under half the cost, which is the kind of trade that actually moves a Claude Code bill. The same roundup and separate testing from Simon Willison both note Opus 5.5 can overthink simple tasks at its maximum reasoning-effort setting, burning tokens for no real gain, so it is worth testing the default effort level before assuming more thinking means a better result.

Has Claude Code Changed Its Pricing Recently?

Yes, and the most notable change this year was one Anthropic announced and then reversed. On May 14, 2026, Anthropic said it would move the Claude Agent SDK and headless (`claude -p`) usage off subscription limits and onto a separate, dollar-denominated metered credit. The change was scheduled to take effect June 15, 2026, and Anthropic paused it on that exact day, stating plainly that “nothing has changed” for now.

Instead of a blanket metering switch, Anthropic has taken a narrower approach: gating specific capabilities, like Opus-class access on the Pro plan or the specialized Fable model line, behind usage credits rather than remetering everything a developer does. Weekly rate limits have also shifted more than once since Claude Code launched, including added peak-hour throttling on weekday mornings introduced in March 2026 and a 20 percent increase to five-hour session limits that came with the September 22, 2026 Opus 5.5 release, so treat any rate-limit figure you read, including the ones in this article, as a snapshot rather than a permanent ceiling.

What Actually Drives Your Claude Code Bill Up?

Four things push spend higher than a casual user expects: long-running sessions that never get cleared, defaulting to Opus for routine tasks, running multiple parallel agent teammates, and cache misses after a break longer than the prompt cache’s lifetime. None of these show up as a single alarming charge; they compound quietly across a session.

Agent teams are the biggest lever most people underestimate. Running several Claude Code instances in parallel to divide up a task uses roughly seven times the tokens of a single standard session when teammates operate in plan mode, since each one maintains its own full context window. A three-person agent team left running overnight can consume more budget than a week of normal solo use.

How Does Claude Code Pricing Compare to Other AI Coding Tools?

Claude Code’s flat-rate subscriptions sit in a similar band to competing agentic coding tools, but the billing mechanics differ enough to change the real cost. Where Claude Code ties usage to five-hour rolling windows shared across chat and code, some competitors meter more narrowly around IDE completions or charge per agent action instead.

If you are choosing between tools rather than just plans, our Claude Code vs Cursor comparison lines up both products’ pricing tiers and usage limits side by side, which matters more than the sticker price alone once you factor in how each tool counts a “request.” Teams evaluating the wider field should also see our roundup of AI tools for developers, since cost-per-seat comparisons only tell half the story without knowing what each tool actually replaces in your workflow.

Is Claude Code Worth the Price for a Small Team?

For most small teams, yes: two Team Standard seats at $40 a month total is cheap relative to even a few hours of saved developer time weekly, and the shared usage pool, admin controls and SSO justify the jump from individual Pro accounts once you have more than one person coding regularly. The calculation changes if your team runs long autonomous agent sessions, where Team Premium or a metered API workspace controls runaway spend better than a flat seat.

The honest failure case, visible across developer forums, is a team that underestimates how usage stacks: one widely discussed report described a Max 20x subscriber whose API-equivalent usage in a single burst month would have cost well over $4,000 had it been billed per token, dramatically more than the $200 flat fee they actually paid. That example cuts in the subscriber’s favor, but it shows the gap between what a heavy session actually consumes and what a flat plan charges can run in either direction depending on which side of the meter you are standing on.

Which Claude Code Plan Should You Actually Pick?

Start with Pro at $20 a month if you code with Claude for an hour or two most days. Move to Max 5x once you are hitting session limits weekly, which the community consistently reports happens within the first month of daily heavy use. Only jump to Max 20x, Team, or a metered API workspace if you are running multi-hour sessions, parallel agent teams, or CI-triggered automation that a flat personal plan cannot absorb.

For teams, the seat-versus-API decision usually comes down to predictability: seats cap your maximum spend and simplify budgeting; API billing scales with actual work done but requires someone watching the console dashboard or setting workspace spend limits before a runaway job surprises finance.

Conclusion

Claude Code pricing is not complicated once you separate the two billing worlds it lives in: a capped subscription seat that shares usage with Claude’s chat product, or an uncapped API meter billed strictly per token. The free tier gets you neither; it gets you chat, and nothing more.

If you are still deciding where Claude Code fits against the rest of your AI toolchain, start with our full AI tools directory for context on how coding assistants stack up against the broader category, then pick the smallest paid plan that matches your actual usage pattern rather than the one with the biggest number attached.

Track your first month with `/usage` before committing to anything above Pro. Real session data beats a marketing tier every time.