Plans, seats & accounts

Why the plan decision comes first

The outcome here is unglamorous and decisive: a working account on the right plan, every seat verified, and a named billing owner - the decisions that determine whether week one flies or dies on login screens. Claude Code is not a separate product you buy: it comes with your Claude subscription, and the plan decides which models you get by default, how much usage you have, and whether you get team admin controls at all.

One thing to get straight on day one: the free Claude plan does not include Claude Code. You need Pro, Max, Team, Enterprise, or an API account through the Claude Console. If you try to log in on a free account, it will not work.

The plans, plainly

• Pro - the solo starting point: $20 a month, or $17 on annual billing, as of September 2026. Includes Claude Code with Opus 5.5 as the default model. Good for learning and light daily use; heavy days will hit the five-hour usage window.
• Max - the solo power plan. Max 5x is $100 a month and Max 20x is $200 (as of September 2026): the same models with 5x or 20x the Pro usage per window. If Claude Code becomes your daily driver, this is where you land.
• Team - the plan for 5-15 person orgs. Self-serve, per seat, minimum two seats, with an admin dashboard. Team Standard is $20 per seat per month annual ($25 monthly); Team Premium is $100 per seat annual ($125 monthly) with about 5x the usage of a Standard seat. Both default to Opus 5.5.
• Enterprise - adds SSO, domain capture, a compliance API, and org-wide model and permission policy. Annual only, priced per seat plus usage; get the number from sales, not from a course.
• Console (API) - pay-as-you-go instead of a subscription. You pay per token. This is how scheduled jobs and CI usually authenticate, separate from anyone's personal seat.

The default model is the same everywhere now: Opus 5.5 on Pro, Max, Team, Enterprise and API accounts, since Claude Code v2.1.280 (September 22, 2026). Older guides that say Sonnet on Pro and Opus on Max are out of date. Fable 5.1, the top model, is never a default: you opt in with /model fable, and on plans where it bills to usage credits Claude Code shows a consent prompt first. You can switch models any time inside a session - week two's 'Models, effort and cost' lesson covers the lineup, the prices, and when each model earns its cost.

Seats and admin for teams

Solo readers can skip to the next section. For everyone else: the admin work below takes twenty minutes and saves the week-one workshop from dying on login screens.

Where usage actually comes from

When you use Claude Code interactively - typing in a session - usage draws from your subscription plan limits. When a script or server calls Claude with an API key, that draws from API credits and is billed per token. Two pools, two bills.

One thing older guides get wrong: Anthropic announced a separate monthly Agent SDK credit for headless runs (claude -p), GitHub Actions and SDK apps starting June 15, 2026, then paused it the same day. As of September 2026 nothing changed: headless and SDK usage on a subscription still draws from your plan's limits, a rolling five-hour window plus a weekly cap, shared across Claude chat and Claude Code. Automations that must not compete with people's sessions run on an API key.

The rule we run internally and install with every client: two wallets. In-session work (classifying a list, summarizing calls, drafting) runs on the subscription seat. The metered API key is only for unattended systems: cron jobs, a deployed service, a feature a client runs. And a key being available is not permission to spend it. Before any batch run, whoever runs it states per-unit cost times unit count, and the billing owner says yes.

What does it cost in practice? Anthropic's own published benchmark: about $13 per developer per active day, roughly $150-250 per person per month for heavy users, and 90% of users stay under $30 a day. For operators doing GTM and ops work, expect the lower end.

Check your setup before moving on

Thirty minutes, four checks. Done honestly, the rest of week one is install mechanics and your first real wins.

1. Decide your plan: Pro (solo, starting), Max (solo, heavy), or Team (5-15 people).
2. Create or upgrade the account at claude.ai using your work email.
3. Teams: invite everyone from the admin dashboard and confirm each seat includes Claude Code.
4. Write down who owns billing and admin - one named person, not 'whoever set it up'.

Do this now

Sources and further reading

• Claude Code documentation
• Usage and cost monitoring docs