Claude Code Billing Options in 2026

Claude Code billing options are a paid Claude plan (Pro, Max, Team, or Enterprise) or API credits from the Anthropic Console. There is no free Claude Code tier.

The subscription shares one usage pool with claude.ai. Credits bill per token at published API rates.

People mix these up because both logins work in the same CLI. The bill is not the same.

What are the claude code billing options right now?

As of August 2026 the menu is short.

Claude Pro

Type: individual subscription.

Price: $20 a month, or about $17 a month billed annually.

Usage: a rolling five-hour window shared with Claude chat. Fine for light terminal work.

You will hit it if you code for hours.

Claude Max 5x

Type: individual subscription.

Price: $100 a month.

Usage: five times Pro per session, plus a weekly cap. This is the default for people who live in the CLI on weekdays.

Claude Max 20x

Type: individual subscription.

Price: $200 a month.

Usage: twenty times Pro per session. For all-day agent loops, not for occasional edits.

Team Standard and Team Premium

Type: per-seat subscription.

Price: about $25 a seat monthly for Standard, about $125 a seat monthly for Premium, cheaper on annual.

Usage: team admin, more headroom than Pro on Premium. Claude Code is included.

See Anthropic's current Team page before you invoice finance.

Enterprise

Type: seat plus usage, sales-quoted.

Price: a seat fee plus API-rate usage on many contracts.

Usage: SSO, compliance, and pooled reporting. The bill looks like infrastructure, not a consumer plan.

API credits (Claude Console)

Type: pay as you go.

Price: published per-million-token rates for the model you picked. No monthly Claude Code fee.

Usage: no plan ceiling unless you set one. Easy to surprise yourself if auto-reload is on.

How to use Claude Code is the product walkthrough. This page is only the bill.

How does subscription usage actually reset?

Session limits reset on a rolling five-hour window. Weekly limits reset on a fixed weekday assigned to your account.

You can see the next reset under Settings, Usage. The day does not move when you subscribe.

Hitting a session cap mid-refactor is the usual reason people jump from Pro to Max. Hitting a weekly cap is the reason they jump from Max 5x to Max 20x.

One pool feeds chat and the terminal. A long web session can starve the CLI.

I treat chat as a second tab on the same meter.

Anthropic has said most Claude Code users stay under a modest per-active-day average when they measure token-billed deployments. Treat that as a published average, not your forecast.

What happens if I set ANTHROPIC_API_KEY?

The CLI prefers the environment key. If ANTHROPIC_API_KEY is set, Claude Code bills the Console at API rates and ignores Pro or Max.

That is the number one surprise I see. A leftover key from an old script turns a $20 plan into a token firehose.

Auto-reload in the Console is a separate switch. It only spends after you have opted into credits.

Decline the credit prompt if you are protecting a monthly budget. The CLI should not silently switch pipes.

Which claude code billing options fit which week?

I use a boring rule.

Occasional CLI, daily chat

Pick Pro. You will feel the five-hour window.

That is the signal to upgrade, not a bug.

Weekday agent loops

Pick Max 5x. $100 is ugly until you compare it with a week of API credits on Opus.

All-day autonomous runs

Pick Max 20x, or credits with a hard cap. If you run overnight loops, a weekly Max cap can still stop you.

A team that needs receipts

Pick Team or Enterprise. Personal Max on a company card becomes a compliance problem by month three.

Hard spend cap, weird hours

Pick Console credits with a limit and no auto-reload. You trade price predictability for a ceiling.

Claude Code versus Cursor is the product comparison. Cursor is a different bill (Cursor subscription plus whatever models it routes).

Do not convert the two line by line.

How do API rates compare to a Max plan?

API rates change by model. Read the live table before you forecast.

A heavy Opus day at API rates can exceed Max 5x by itself. A light Sonnet day can undercut Pro.

Anthropic has published averages around the low tens of dollars per active developer day for token-billed rollouts. Twenty such days is already more than $100.

Subscription wins for daily habit. Credits win for burst work, evals, and CI jobs you can turn off.

CI is the clean credits case. A GitHub Action should not sit on your Max plan.

It should sit on a Console key with a tiny cap.

Local MCP tools do not change the token bill much if they return short structured results. They change it a lot if you paste a 4K screenshot into every turn.

Claude Code MCP setup is how you add local tools. A named element plus a crop is cheaper than a full-window PNG.

How should I keep Claude Code cheap without starving it?

Give the model less pixel junk.

A screenshot of a whole desktop is a token dump. A crop of one control plus its AX role, label, frame, and confidence is a short tool result.

PinVari captures that on-device. Hold Option-Command-A, circle the control, speak.

The local MCP connector on 127.0.0.1:3402 hands Claude Code pinvari_next_instruction instead of a gallery of PNGs.

Connect with one click inside the app, or:

claude mcp add --scope user pinvari -- "$HOME/.pinvari/mcp/pinvari-mcp"

The app must be installed and running. Bare claude mcp add pinvari errors.

On-device speech and OCR stay on the Mac. You still pay Anthropic for the model.

You do not pay a second cloud to look at the screen.

If you want the capture layer as a one-time Mac licence, see pricing.

What does Team billing change?

Team seats include Claude Code. Admins see usage.

You can keep people off personal Max cards.

Premium seats exist because Standard seats still run out. Budget the seat you will actually use, not the cheaper line item.

Enterprise often adds usage at API rates on top of a seat. Ask for the unit economics in writing.

A "cheap seat" plus uncapped usage is a credits plan in a blazer.

SSO and retention policies are the real Enterprise buy. Do not buy Enterprise to save money on tokens.

Finance wants a SKU. Give them Pro, Max, Team, or Console credits.

Do not invent a fourth "Claude Code plan." There isn't one.

How do I switch billing options without losing the session?

You can log out of the CLI and log in with the other account. Running sessions die.

That is fine.

You cannot merge a Pro remainder into credits. Unused session capacity expires with the window.

If you switch to credits mid-month, cancel or keep the subscription on purpose. Leaving both on is how people double-pay.

If you switch from credits to Max, remove the env key first. I say that twice because I have paid for the mistake.

Pause/resume in your capture tool (Option-Command-P in PinVari) does not pause the model bill. Only the CLI usage window and the Console meter do that.

What should I tell a PM who only sees the sticker?

Pro is the trial that includes the real CLI. Max is the working developer's plan.

Credits are for machines and spikes.

Claude Code is not on Free. A teammate on Free can install the binary and still fail login.

Open-source maintainers sometimes get limited Max grants. Treat those as time-boxed, not as a strategy.

The model you pick inside the CLI (Sonnet versus Opus) moves the credit bill more than it moves a Max sticker. On a subscription, a bigger model burns the window faster.

Pick the model for the task, then pick the bill that survives that model.

Wispr Flow at $12 to $15 a month proves people pay for push-to-talk on a Mac. That is a separate bill from Claude.

PinVari is a one-time Mac app that aims the same habit at UI feedback.

Jam.dev is free for browser bugs. It does not replace Claude Code and it does not appear on this invoice.

FAQ

Does Claude Pro include Claude Code?

Yes. Every paid Claude plan includes Claude Code and shares the plan's usage pool.

Free does not include it. You still have to log the CLI into that plan.

Are Max 5x and Max 20x the only heavy claude code billing options?

No. You can also burn Console credits at API rates, or sit on Team Premium / Enterprise.

Max is the simple individual upgrade. Credits are the uncapped alternative.

Why was I charged API rates on a Max plan?

ANTHROPIC_API_KEY was set, or you accepted the credit prompt after a limit. Unset the key, log in again with the plan, and decline credits if you want the subscription only.

Can I use Claude Code only with API credits and no subscription?

Yes. Log the CLI into a Console account, buy credits, and optionally disable auto-reload.

You pay token rates with no Pro or Max fee.

Do MCP tools change Claude Code billing?

They change token volume, not the price list. Short structured tool results (a named element, a path, a crop) cost less than pasting screenshots.

The plan or the credit meter still pays for the model.

Is there a student discount on Claude Code itself?

Anthropic sometimes discounts consumer Claude plans. There is no separate student SKU for the CLI.

Check the live Claude pricing page, not a third-party screenshot.