Claude Code pricing in 2026: plans, limits, and API costs

Claude Code pricing is simpler than the search results make it look. There's no Claude Code plan. You buy a Claude plan, and Claude Code comes with it.

The catch is that every plan meters usage. Pro, Max, and Team all run the same agent in your terminal. What changes is how long you can keep it busy before you hit a wall. We checked every number below against Anthropic's own pages on 8 October 2026, including a help-center change that went live the same day.

How much does Claude Code cost on each plan?

Here's what Anthropic lists for the plans that include Claude Code. Prices are USD before tax.

Sources: Claude pricing, Team pricing, and the Max plan help article. Team plans cover 2 to 150 people.

One quirk worth knowing: the main pricing page says Max is "from $100." The two tiers, 5x and 20x, are spelled out in the help center. Max has no annual discount.

What do you actually get for the money?

A share of compute, measured in sessions. Anthropic doesn't publish token counts for each plan, so nobody can tell you "Pro gets X messages" with a straight face.

Here's what Anthropic does say:

• One pool for everything. Claude Code and the Claude apps draw from the same allowance. A long refactor in the terminal leaves less for chat that afternoon.
• 5-hour rolling windows. Your session allowance resets on a rolling 5-hour clock. Paid plans also have weekly limits on top.
• Relative sizes. Pro gives at least 5 times the Free allowance per session. Max 5x and Max 20x give 5 and 20 times Pro.
• Fable is rationed. On Pro, Fable 5.1 only runs on paid usage credits. Max and Team Premium can spend up to half of their weekly limits on it.
• Extra usage costs API rates. When you run out, you can buy usage credits and keep going. Those are billed at standard API prices.

The model you pick matters as much as the plan. If you leave Claude Code on Fable at High effort, a Max 5x pool drains faster than it would on Sonnet. Our breakdown of Claude Fable effort levels and cost covers that trade.

What does Claude Code cost on the API instead?

If you sign in with a Claude Console API key, you skip the plans and pay per token. These are the current rates per million tokens (MTok):

Source: Claude pricing, checked 8 Oct 2026. US-only inference costs 1.1x these rates.

A quick sense check on Sonnet 5.5: 1 million input tokens and 100,000 output tokens cost $2 plus $1, so $3. Agent sessions read a lot of files, which is why cache reads matter. A cache read on Sonnet is $0.10 per MTok, a twentieth of a fresh input token.

What does Anthropic say a typical developer spends?

Anthropic's costs page gives enterprise averages: about $13 per developer per active day, $150 to $250 per developer per month, and under $30 per active day for 90% of users.

Those are Anthropic's numbers from enterprise deployments. We haven't measured our own spend against them, so treat them as a reference point.

Is a subscription or the API cheaper for Claude Code?

For anyone coding with Claude most weekdays, a subscription usually wins. Max 5x at $100 sits below the bottom of Anthropic's $150 to $250 monthly API range. Pro at $20 is far below it, if its limits hold up for your work.

The API wins in two cases:

1. You use Claude Code a few days a month. Paying $13 on the days you work beats $100 for a month you mostly skip.
2. You run it unattended. Scripts, CI jobs, and headless runs with claude -p are easier to budget per token, and they don't burn the pool you need for interactive work.

There's one more difference people miss. Anthropic's costs page says prompt caches last 1 hour on a subscription and 5 minutes on the API or on usage credits. Long, start-and-stop sessions get more cache hits on a plan.

What changed on 8 October 2026: monthly API credits for Max and Team

Anthropic published a help article on 8 October 2026 that adds monthly Claude Platform credits to Max and Team.

Team credits pool into one balance capped at $500 a month. Free, Pro, and Enterprise aren't eligible. You need 7 days on the plan, and you link one Claude Console organization to receive the credit.

Read the fine print before you count this as free Claude Code:

• The credits don't cover interactive Claude Code in the terminal, IDE, desktop, or web. They don't cover extra usage after you hit limits either.
• They do cover the Claude API, the Console Playground, Managed Agents, the Agent SDK, and claude -p run with an API key from the linked organization.
• They don't roll over, and they're spent before any credits you bought.

So a Max 20x subscriber now gets $200 of API spend each month for side projects, scripts, and agents, alongside the same Claude Code limits as before. Source: Monthly API credits for Max and Team plans.

How do you keep Claude Code costs down?

These come from Anthropic's Manage costs effectively page:

• Run /usage to see where your allowance went.
• Use /clear between unrelated tasks, and /compact when a long session bloats.
• Default to Sonnet. Save Opus and Fable for the hard parts.
• Turn effort down with /effort for routine edits.
• Be careful with agent teams. Anthropic says they use roughly 7 times more tokens than a single session.

Background usage (summaries, small helper calls) is typically under $0.04 per session, per the same page. The big spend comes from model choice and context size.

Which Claude Code plan should a solo builder pick?

Start on Pro. It's the cheapest way to find out how much Claude Code you actually use, and annual billing drops it to $17.

Move to Max 5x when the 5-hour limit is stopping real work more than a couple of times a week. That's our rule of thumb, and you'll know when you hit it. Max 20x is for people who run Claude Code most of the day, or who want headroom for Fable.

Pick Team when two or more people need central billing and admin controls. A Premium seat costs the same as Max 5x on annual billing, and it gets 5 times the usage of a Standard seat.

Still deciding between tools? Our Cursor vs Claude Code comparison covers the editor side, and ChatGPT Plus vs Claude Pro vs Google AI Pro compares the $20 tier across vendors.

What we couldn't verify

• Exact limits. Anthropic doesn't publish message or token counts per 5-hour window. Any number you see for that comes from user reports.
• Our own spend. We haven't published MorningStacks usage data here, so the cost figures above are Anthropic's.
• Regional prices. Everything here is USD. Local pricing and taxes vary by country.