Claude Code pricing: what it actually costs to use
Claude Code isn't in the Free plan. Pro at $20/month is the entry point and covers a beginner's first project; Max and API billing exist for heavier use.
The short answer
Claude Code doesn't have a free tier. The cheapest way in is Claude Pro at $20/month ($17/month billed annually), which includes a usage pool shared with Claude.ai chat. That's enough to learn Claude Code and build a first website or small app. Heavier users move to Max ($100 or $200/month) for a bigger weekly allowance, or pay per token through the API with no subscription at all.
This site isn't affiliated with Anthropic. We read the pricing and support pages directly and link to them below.
Which plans include Claude Code
| Plan | Price | Claude Code included? | Who it's for |
|---|---|---|---|
| Free | $0 | No | Trying Claude.ai chat only |
| Pro | $20/month ($17/month billed annually) | Yes | Individuals learning Claude Code or building a first project |
| Max 5x | $100/month | Yes | Daily coding sessions that outgrow Pro's limit |
| Max 20x | $200/mo | Yes | Heavy, near-constant use or long agentic sessions |
| Team Standard | $25/seat/month ($20/seat/month billed annually) | Yes | Small teams that want one bill and shared admin controls |
| Team Premium | $125/seat/month ($100/seat/month billed annually) | Yes | Teams that need heavier per-seat usage than Standard |
| Enterprise | $20 /seat + metered usage; sales-assisted options available | Yes | Large orgs needing SSO, audit logs, spend controls |
The Free plan explicitly excludes it: Anthropic's own help article on the subject "applies to individual consumers using Pro or Max plan subscriptions to access Claude Code," with no Free-tier path mentioned in it (Use Claude Code with your Pro or Max plan).
The figures above are Anthropic's US-dollar prices. Consumer checkout pricing and taxes can vary by region, while the Claude Platform pricing documentation lists API rates in USD. Check your local upgrade screen before treating a converted figure as a quote. For the full breakdown of what each Claude plan includes beyond Claude Code, see Claude pricing.
| Cheapest plan with Claude Code | Pro, $20/mo |
|---|---|
| Free plan | No Claude Code access |
| Max 5x | $100/mo, 5x Pro's usage |
| Max 20x | $200/mo, 20x Pro's usage |
| API pay-as-you-go | No subscription, billed per token |
How the usage limits actually work
Claude Code doesn't get its own separate quota from Claude.ai chat on Pro and Max. Anthropic is explicit about this: "usage limits are shared across Claude and Claude Code, meaning all activity in both tools counts against the same usage limits" (Use Claude Code with your Pro or Max plan). Spend the morning chatting on claude.ai and the afternoon coding, and both draw from one pool.
On top of that shared pool, usage resets on a rolling session window rather than a fixed daily clock, so how much room you have depends on when your current session started, not the time of day. Anthropic describes usage as budget-based rather than message-counted, factoring in conversation length, model choice, and effort level rather than a flat number of prompts.
Max plans add a second layer. Anthropic's own Max plan article states there are "two weekly usage limits: one that applies across all models and another for Sonnet models only," and that "weekly limits reset at a fixed time assigned to your account" (What is the Max plan?). That means you can run out of Sonnet-specific weekly budget even while your overall weekly allowance still has room, since the two caps track separately.
Run /model inside Claude Code to see what's available and which model is active. Sonnet is the default, and Anthropic recommends keeping it that way: "the right choice for the large majority of coding work. It is fast, capable, and cost-efficient." Opus handles harder problems like large refactors, but "uses meaningfully more of your quota" (Models, usage, and limits in Claude Code). For what these limits mean on Claude.ai chat specifically, see Claude usage limits.
The API alternative, and the credits trap
You can also run Claude Code without any Pro or Max subscription, authenticated with an API key instead. That mode "charges by API token consumption" (Manage costs effectively) with no included quota, and critically, has "no hard stop" on spend: the account is charged for exactly what it uses (Models, usage, and limits in Claude Code).
Current per-token rates (per million tokens, input/output): Claude Haiku 4.5 runs $1/$5, Claude Sonnet 5 is $2/$10 through an introductory window ending August 31, 2026 then $3/$15, and Claude Opus 4.8 is $5/$25.
Prompt caching cuts a cache-hit read to a tenth of the base input price, but TTL and authentication change what is actually reused; the Claude Code prompt caching guide explains those conditions. The Batch API halves both input and output for jobs that don't need to run immediately (Anthropic pricing documentation).
The trap isn't a dollar figure, it's that API-key usage is metered rather than included in a subscription allowance. An unattended session can run longer than you meant it to unless you configure workspace or provider limits. Pro and Max users can set a monthly spend limit on usage credits with the /usage-credits command, and Claude Code asks before moving beyond that limit (same costs documentation).
Anthropic's own enterprise usage data is a useful sanity check against any horror story you've seen online: on API billing, the average runs about $13 per developer per active day, or $150-250 per developer per month, with under $30/active-day covering 90% of users, and background token use between prompts typically under $0.04 per session (Manage costs effectively). Most usage is modest; the expensive outliers are multi-agent setups or long unattended runs, not ordinary day-to-day coding.
Cost-saving habits worth building early
Anthropic's own cost guidance boils down to a handful of habits that cost nothing to adopt:
Use Plan Mode for anything nontrivial
Shift+Tab to enter plan mode before a big change. Anthropic calls this a major lever against expensive rework, since Claude reasons through the approach before touching files.
Run /clear between unrelated tasks
Anthropic describes this as the single most effective habit for both quality and cost. A long, accumulated conversation costs more per turn than a fresh one.
Keep Sonnet as your default
Reserve Opus for genuinely hard problems: big refactors, tricky debugging, architectural decisions. Switching models mid-session is one command away.
Trim what loads automatically
Disable MCP servers you're not using via /mcp, and keep CLAUDE.md short. Workflow instructions that only apply sometimes belong in a skill that loads on demand, not a file read on every turn.
Frequently asked questions
Is Claude Code free?
No. There is no Free-tier path to Claude Code. The cheapest plan that includes it is Pro at $20/month ($17/month billed annually). Anthropic's own support documentation frames Claude Code access specifically around Pro and Max subscriptions, with no mention of Free-tier eligibility.
Is Claude Pro enough for Claude Code?
For most beginners, yes. Pro's usage pool is shared with Claude.ai chat, but Anthropic's own cost guidance targets exactly this user: someone writing a first website or small app, using Plan Mode, clearing context between tasks, and sticking with the default Sonnet model. Long unattended sessions or multiple parallel agents will likely outgrow it.
Why did Claude Code stop working mid-week?
Check the limit message or run /usage: you may have hit the five-hour session limit or a weekly limit. Max plans have two separate weekly caps, one across all models and one just for Sonnet, and either one running out can interrupt you even if the other still has room. Wait for the indicated reset or turn on usage credits with /usage-credits.
Should you use API billing instead of a subscription?
Only for a specific reason, like running Claude Code inside an automated pipeline where a flat monthly plan doesn't fit. For everyday coding, Pro or Max gives a predictable monthly cost; API billing has no included quota and no hard stop, so set a spend limit before you start if you go this route.
If you're brand new to Claude Code and don't write code for a living, Claude Code for non-devs is the place to start before worrying about pricing at all. And for a full breakdown of every Claude plan beyond just Claude Code, see Claude pricing.