Claude Code · September 2026
Claude Code on the Max plan
What $100 or $200 a month buys, how the limits are actually measured, and which tier fits the way you run agents.
What the Max plan is.
Max is Anthropic’s individual plan above Pro, sold in two tiers. Max 5x is $100 a month and gives five times the per-session usage of Pro; Max 20x is $200 a month and gives twenty times. Both include Claude Code alongside the Claude apps under one subscription, with usage limits shared across everything — a long Claude Code session and an afternoon of chat draw from the same pool. Max also carries priority access to new models and features.
The model defaults differ by plan, per Anthropic’s model configuration docs: on Max the default model is Opus, while Pro and Team standard seats default to Sonnet. Opus with the 1M-token context window is included on Max and requires usage credits on Pro. Either tier can switch models mid-session with /model.
How the limits are measured.
Two clocks run at once. The session limit resets every five hours; the weekly limit applies across all models and resets on a fixed day assigned to your account, and the usage screen also shows a separate weekly reset for Opus. What counts is the work, not the message count: conversation length, tool use, model choice, and effort level all draw on the allowance, and Claude Code sends the whole conversation with every request, so a one-line question in a session that has been open all day still costs the whole history. Prompt caching softens that on a subscription — the cache lives an hour — which is why the first message after a long break is the expensive one.
When you reach either limit, Claude Code tells you which and when it resets. A model-specific message such as “You’ve hit your Opus limit” means switching to another family with /model keeps you working; a session or weekly limit does not. From v2.1.234 Claude Code can wait and continue the interrupted task on its own after the reset. Beyond that, Pro and Max can turn on usage credits, which bill anything past the included allowance at standard API rates under a monthly spend cap you set.
Subscription against the API, for Claude Code.
Prices from claude.com/pricing and Anthropic’s support articles on 2026-09-03.
| Plan | Price | Allowance | Default model | Past the limit |
|---|---|---|---|---|
| Pro | $20 a month, or $17 billed annually | Baseline five-hour and weekly limits | Sonnet | Usage credits at API rates |
| Max 5x | $100 a month | 5x Pro per session | Opus | Usage credits at API rates |
| Max 20x | $200 a month | 20x Pro per session | Opus | Usage credits at API rates |
| Console (API key) | Per token, no monthly fee | No session window; organization rate limits | Your choice | It keeps metering |
Anthropic’s own figure for API-billed teams is around $13 per developer per active day and $150–250 a month — more than Max 5x for a full month of daily use.
Which tier, honestly.
The tiers are multiples of the same allowance, so the question is how often you hit the wall, not what you do.
Stay on Pro
An hour or two of Claude Code a day, one session at a time, Sonnet doing most of the work. If the five-hour limit interrupts you less than a couple of times a week, the upgrade buys nothing.
Max 5x
The daily driver. Opus by default, long refactors, one or two sessions open, and the limit stops being something you plan around.
Max 20x
Several sessions in parallel — worktrees, subagents, an agent that runs most of the day. Per session it is the same product as 5x; what you are buying is the room to run several at once. Here the weekly limit is the number to watch, since parallel sessions spend it in parallel.
Seeing what you have used.
Inside Claude Code, /usage (its alias is /cost) shows the plan usage bars for the session and the week, and a breakdown of what drew on them — skills, subagents, MCP servers, long context, cache misses. The Session block at the top prices your tokens at API list rates; Anthropic’s cost docs note it is for API users and not relevant to a subscriber’s bill. The figures come from this machine’s history, so other devices and claude.ai are not included; Settings > Usage on claude.ai has the account-wide view and the next reset time.
/status answers the other question: what you are paying with. Its Login method row shows the subscription account, and an API key row appears when a key is in use. If ANTHROPIC_API_KEY is set in your environment, the key outranks your subscription once approved and every token is metered — unset ANTHROPIC_API_KEY puts you back on the plan.
Making the plan go further.
The levers are all about context. /clear between unrelated tasks, because stale history is resent on every turn; /compact with instructions when the history matters; Sonnet for routine work and Opus for the hard decisions, switched with /model; a lower /effort for simple tasks, since thinking tokens are billed as output. Keep CLAUDE.md under 200 lines and move workflow detail into skills that load on demand; disable MCP servers you are not using; push verbose operations — test runs, log reads — into subagents so only a summary comes back. On Pro and Max, resuming a large session after an hour offers a summary instead of the full history, and taking it is usually right.
One lever is structural. The plan is per account, not per machine: sign in on the workstation, the build box, and the laptop and all three draw on the same allowance. With spawnd that is the whole billing story — the Claude Code CLI logs in on each host as it always does, so one subscription serves every machine you own from any device, with no API key in the middle and no markup on top. spawnd holds no provider credentials; your plan is used exactly as the vendor’s CLI uses it. If you are choosing between tiers, the plan calculator does the arithmetic; if you are asking whether you need a plan at all, start with is Claude Code free.
Start
Same plan, every host you own.
curl -fsSL https://spawnd.dev/install.sh | shQuestions
- Does Claude Code charge my Max plan or the API?
- Your plan, when you signed in through claude.ai. If the ANTHROPIC_API_KEY variable is set and approved it takes precedence and you pay per token; the /status command shows which is active.
- Can I use one Max plan on several machines?
- Yes. Limits are per account and shared across every device and the Claude apps, so three signed-in machines draw on one allowance. Anthropic’s advertised limits assume ordinary individual use; sharing an account between people is what Team seats are for.
- How do Team seats compare to Max?
- A Team standard seat is 1.25x Pro per session and a premium seat 6.25x, so a premium seat sits just above Max 5x. Team plans need at least two members and add admin controls and spend limits.