Claude Code Pricing in 2026: What You'll Really Pay
Garrett Pierson
Claude Code costs $20/month on the Claude Pro plan, $100 or $200/month on Max, or pay-per-token if you bill through the API. Claude’s free plan exists, but it doesn’t include Claude Code — Anthropic’s docs state plainly that “the free Claude.ai plan does not include Claude Code access.” For one founder building one app, Pro at $20 is usually the right answer.
If you’ve searched this and come away scared by threads about people burning $100 a day, those numbers are real and they’re describing a different billing model than the one you’d be on.
Claude Code pricing at a glance
Every plan that includes Claude Code, taken from Anthropic’s pricing page as of August 2026:
| Plan | Price | What you get |
|---|---|---|
| Free | $0 | No Claude Code access at all |
| Pro | $20/mo ($17/mo billed annually, $200 up front) | Claude Code included, standard usage limits |
| Max 5x | $100/mo | 5× Pro’s capacity per session |
| Max 20x | $200/mo | 20× Pro’s capacity per session |
| Team | $20/seat/mo billed annually ($25 monthly) for a Standard seat; $100 ($125) for Premium | Claude Code on every seat; Premium carries 5× a Standard seat’s usage |
| Enterprise | $20/seat plus usage billed at API rates | No per-seat usage cap on usage-based plans |
| API (Console) | Pay per token | No flat fee, metered by usage |
Claude Code is included in every paid plan. Moving from Pro to Max mainly buys capacity — Max also carries higher output limits and earlier access to some features, but the tool you use is the same one.
The two billing models people mix up
Claude Code looks wildly expensive in some forum threads and cheap in others because people are describing two different billing models.
Subscription (Pro, Max, Team). You pay a set monthly fee, and usage is included up to your plan’s limits, which reset on a rolling five-hour window and a weekly window. When you hit a limit, you wait for the window to reset or you upgrade. One caveat worth knowing before you’re surprised by it: usage credits let you keep working past your plan’s limit, billed at API rates. Leave them off and your monthly cost stays fixed at the sticker price.
API / Console. You attach a Console account and pay per token consumed. You can cap it with workspace spend limits, but nothing caps it by default. This is the billing model behind most of the alarming figures you’ll find online.
Anthropic publishes the real API numbers in its cost management docs, and they’re worth reading carefully: across enterprise deployments, “the average cost is around $13 per developer per active day and $150-250 per developer per month, with costs remaining below $30 per active day for 90% of users.”
Those figures describe API-billed enterprise deployments, measured per developer per active day. A founder building one app in the evenings on a $20 subscription is on a different meter, with a fixed ceiling.
What token pricing actually looks like
If you do go the API route, you’re billed per million tokens, and the model you pick moves the number a lot. Current published rates:
| Model | Input (per million tokens) | Output (per million tokens) |
|---|---|---|
| Haiku 4.5 | $1 | $5 |
| Sonnet 5 | $2 | $10 |
| Opus 5 | $5 | $25 |
One dated figure to watch: Sonnet 5’s $2/$10 is introductory pricing that runs through August 31, 2026. From September 1 it goes to $3/$15. If you’re reading this after that date, add 50% to the Sonnet row.
Cached content reads at 10% of the input price. Claude Code re-sends your conversation with every request, so caching is what keeps a long session from costing linearly more. A long session still costs more than a short one.
What $20 actually buys you
On Pro you get the full tool: the desktop app, the terminal version, the VS Code and JetBrains extensions, plan mode, diff review, and CLAUDE.md project rules. Nothing is feature-gated behind Max.
What you’re buying at $100 or $200 is room to work longer without stopping. For a full walkthrough of the tool itself, see how to use Claude Code.
When $20 stops being enough
One signal should drive the decision: you’re hitting your usage limit before you’re done for the day, more than once a week. Claude Code shows a message when you reach a session or weekly limit, along with the time the window resets.
Run /usage inside Claude Code to see where you stand. On a paid plan it shows your plan limits as usage bars, breaks down what’s consuming them, and flags behaviors like long context or cache misses when one accounts for 10% or more of your recent usage. Those figures come from local session history on that machine, so work you did on another device won’t appear.
Three patterns that mean you’ve genuinely outgrown Pro:
- You code most days for hours at a stretch, rather than in evening bursts.
- You run multiple sessions in parallel across different projects.
- You’re working in a large existing codebase where every request carries a lot of context.
If none of those describe you, upgrading mostly buys headroom you won’t reach. A founder shipping their first app usually stays on Pro.
Five ways to spend less
These apply on any plan, and on a subscription they translate to hitting your limits less often:
- Clear between unrelated tasks.
/clearstarts fresh. Stale context rides along on every subsequent message and costs you on all of them. - Match the model to the job. Sonnet handles most coding well and costs less than Opus. Use
/modelto switch, and save Opus for hard architectural problems. We compare the tradeoffs in the best LLM for coding. - Write specific prompts. “Improve this codebase” triggers a broad scan. “Add input validation to the login form” sends Claude straight to the file that needs it. The prompt templates we publish are built around this.
- Use plan mode before big changes. Reviewing a plan costs a fraction of rebuilding after Claude committed to the wrong approach.
- Keep CLAUDE.md short. It loads into context at the start of every session, so every line you add is paid for in each session that follows. Anthropic suggests keeping it under 200 lines.
Is it worth it?
Compare it to the alternative rather than to zero. Hiring a developer to build a basic app runs tens of thousands of dollars, and the meter starts over for every change you want afterward. Claude Code at $240 a year on monthly billing — or $200 if you pay for the year up front — is a rounding error against that, and time is the scarcer resource anyway.
The fair criticism of the $20 plan is the one developers make: if you live inside the tool professionally, you will hit the limits. That criticism is accurate and it doesn’t describe most people reading this.
The bottom line
Claude Code costs $20/month on the plan most non-technical founders should be on, $100 or $200 once you’re in it daily, and per-token if you deliberately choose the API. The free plan doesn’t include it. Start on Pro, run /usage when you feel constrained, and upgrade when you hit real limits on real workdays. The tool is the cheap part of building a business; the rest is what Software Secrets 2.0 covers, and the whole book is free.