Claude Code is available on Pro ($20/mo), Max 5x ($100/mo), Max 20x ($200/mo), Team Standard and Premium seats, Enterprise, and the per-token API. Every paid plan gives the same models and features and differs only in usage allowance and billing shape. Priced per unit of allowance, every tier lands on the same line except Max 20x, which is exactly twice the value per dollar. Usage credits, added in 2026, let you overflow past any subscription tier at API rates instead of upgrading.
- There is no free Claude Code plan. Pro at $20/mo is the floor.
- Every paid plan has identical capability. You are buying allowance, not features.
- Pro, Max 5x, and both Team seats are priced at the same allowance per dollar. Max 20x is double.
- Usage credits now let you overflow past any tier at standard API rates, so upgrading is no longer the only way to keep working.
- Extended 1M context is included on Max, Team, and Enterprise for Opus, and needs usage credits on Pro.
- The API is still the only option for CI, automation, and anything you ship.
Every plan, side by side
Claude Code plan comparison, checked 7 August 2026. Session multipliers are usage per five-hour window relative to Pro.
| Free | Pro | Max 5x | Max 20x | Team Standard | Team Premium | API | |
|---|---|---|---|---|---|---|---|
| Price | $0 | $20/mo | $100/mo | $200/mo | $25/seat/mo | $125/seat/mo | Per token |
| Annual price | n/a | $17/mo | Monthly | Monthly | $20/seat/mo | $100/seat/mo | n/a |
| Claude Code | No | Yes | Yes | Yes | Yes | Yes | Yes |
| Session allowance | Chat only | Baseline | 5x Pro | 20x Pro | 1.25x Pro | 6.25x Pro | Unmetered by plan |
| Weekly limits | n/a | All models | All models + Sonnet | All models + Sonnet | All models | All models + Sonnet | None |
| Opus at 1M context | n/a | Usage credits | Included | Included | Included | Included | Full access |
| Usage credits | No | Yes | Yes | Yes | Admin enables | Admin enables | n/a |
| Runs in CI | No | No | No | No | No | No | Yes |
| Bill can surprise you | No | Only on credits | Only on credits | Only on credits | Only on credits | Only on credits | Yes |
Team plans run from 2 to 150 seats, and both seat types carry Claude Code. Usage limits on Team are per member rather than pooled across the team, so a heavy engineer cannot consume a quiet colleague allowance.
The arithmetic nobody publishes
Anthropic publishes the multipliers and publishes the prices, but never divides one by the other. Doing it yourself produces a strikingly flat line with one exception.
Session allowance per dollar per month, normalized to Pro. Monthly billing unless stated.
| Plan | Price | Session allowance | Allowance per $ |
|---|---|---|---|
| Pro | $20 | 1x | 0.050 |
| Pro, annual | $17 | 1x | 0.059 |
| Max 5x | $100 | 5x | 0.050 |
| Max 20x | $200 | 20x | 0.100 |
| Team Standard | $25 | 1.25x | 0.050 |
| Team Standard, annual | $20 | 1.25x | 0.063 |
| Team Premium | $125 | 6.25x | 0.050 |
| Team Premium, annual | $100 | 6.25x | 0.063 |
Every tier is priced at exactly 0.050 units of Pro allowance per dollar, except Max 20x at 0.100. Annual billing moves the number by 15 to 20 percent. Nothing else on the ladder varies at all, which is a deliberate and quite honest pricing structure: you are not being punished for starting small.
Three questions that pick your plan
Does a program need to run this, or a person?
If a pipeline, a cron job, or a product you ship is doing the calling, the answer is the API and nothing else on this page applies. Subscriptions are licensed for interactive use by one human, and the same restriction is why Claude Code on a plan cannot run in CI.
How many hours a day do you actually have an agent working?
Under an hour: Pro. One to four hours: Max 5x. More than that, or several at once: Max 20x. This single question resolves most individual cases correctly, and you can check the answer rather than guess it by running /usage and pressing w for the last seven days.
Do you run sessions in parallel?
Two concurrent sessions roughly double consumption, and agent teams use around seven times the tokens of a standard session when teammates run in plan mode. Parallelism moves people up a tier faster than anything else, and it is the clearest signal for Max 20x.
Usage credits changed the upgrade decision
Until 2026 the ladder was rigid: hit a weekly cap and your options were wait or upgrade. Usage credits added a third rung. Pro, Max 5x, and Max 20x subscribers can prepay a balance and keep working past the included allowance at standard API rates, with a monthly spend cap, optional auto-reload, and alerts.
That turns tier choice into a straightforward comparison. Ask what your overflow actually costs.
When to upgrade instead of buying credits.
| Your situation | Cheaper option |
|---|---|
| You overflow once or twice a month, briefly | Stay put. Credits |
| You overflow most weeks and credits run past $100/mo | Upgrade Max 5x to Max 20x |
| You overflow on Pro every week | Upgrade to Max 5x. Credits at Pro scale get expensive fast |
| Your overflow is one heavy project this month | Credits, or one month of the higher tier |
| Your overflow is a scheduled or scripted job | Neither. Use an API key |
On Team and Enterprise, credits are an admin decision. An admin turns them on and sets spend limits at the organization, group, or individual member level, and members without billing access run /usage-credits to send a request instead.
The plan people pick wrongly
- An API key for daily interactive work. Still the most common and most expensive mistake, usually made by people who arrived through the developer docs rather than the plan page. The subscription is priced for humans at keyboards and the API for programs.
- Max 20x from day one. Bought as insurance, then never approached. Start at 5x, run
/usagefor a month, and let your own interruption rate decide. - Pro while running Opus by default. Produces constant interruption and the conclusion that Claude Code is unusable, when the fix was one line in
/model. - A Team plan bought as a discount. Per seat it costs more than everyone buying Pro. You are paying for provisioning, policy, and clean offboarding.
- Sharing one account. Windows are per account, so two people on one login halve each other constantly and both have a worse experience than two Pro subscriptions would give them.
Questions people ask
Pro at $20/mo, Max 5x at $100/mo, Max 20x at $200/mo, Team Standard at $25 per seat per month, Team Premium at $125 per seat per month, Enterprise on a custom quote, and the per-token API. There is no free Claude Code tier.
Under an hour of agent work a day, Pro. One to four hours, Max 5x. More than that or parallel sessions, Max 20x. A team that needs central billing and offboarding, Team seats. Anything programmatic, the API.
Per unit of allowance, yes, and by exactly a factor of two: 0.100 units of Pro allowance per dollar against 0.050. In practice that only matters if you would actually exceed Max 5x, because unused allowance is worth nothing.
No. Every paid tier gets every model. The tiers differ in how much you can run, not what you can run. The one capability that is plan-gated is extended 1M context for Opus, which is included on Max, Team, and Enterprise and needs usage credits on Pro.
Yes, on both seat types as of August 2026. A Standard seat carries 1.25 times Pro usage per session and a Premium seat carries 6.25 times, and limits are per member rather than pooled across the team.
Yes. Subscriptions are monthly and you can move up or down, which is why starting lower and upgrading on evidence is the sensible approach. Pro is the only tier normally sold annually.
No. The Claude free tier is chat only, with a session limit that resets every five hours. Claude Code starts at Pro. The cheapest way to try Claude Code without a subscription is an API key with a console budget limit.
You see a session or weekly limit message with the reset time. Session and weekly windows are shared across all models, so switching model does not restore access. If usage credits are enabled you can continue at standard API rates instead of waiting.
Sources
Every figure above was read from these pages on August 2026. Vendors reprice without notice; if you find a stale number, tell us.