← Back to Blog
Is Claude Code Free? What You Pay in 2026.
Claude Code is not free. You need a Pro, Max, Team or API account. Here's what each tier costs in 2026 and the cheapest way to actually use it.
Claude Code is not free. You need a Pro, Max, Team or API account. Here's what each tier costs in 2026 and the cheapest way to actually use it.
Claude Code is not free. As of May 2026, you need a paid Anthropic plan to run it, starting at Pro for $20 a month, or you bring your own API key and pay per token. The free Claude.ai chat plan does not include Claude Code access, and Anthropic's own setup docs say so directly.
If you read an article from 2025 saying Claude Code was free with the Pro plan, that's still mostly true, but the messaging got messy in April 2026 when Anthropic ran a test that briefly removed Claude Code from Pro for a slice of new users. The Register covered it on April 22, 2026. The test was reverted. Existing Pro subscribers were never affected. But the search results haven't caught up. I'm Tom. I've been on a paid Anthropic plan since Claude Code launched, I've shipped real client work on Pro, Max 5x and Max 20x, and I've watched my own usage spike and crash with the rate limits. Here's exactly what you pay, what you get on each tier, and which path is actually cheapest for the way you work.
No. Claude Code requires a paid plan. Anthropic's setup documentation states it plainly: "Claude Code requires a Pro, Max, Team, Enterprise, or Console account. The free Claude.ai plan does not include Claude Code access." The cheapest legitimate route is Pro at $20 a month, or $17 a month if you pay annually. The other route is a pay-as-you-go Anthropic API key, billed per token through the Console. There is no free tier, no free trial that lasts beyond your first session, and no "free with limits" mode like ChatGPT's free chat.
If you saw a screenshot from earlier this year showing Claude Code as a Pro feature, that screenshot is correct again today. The April 2026 test that pulled it for a small percentage of new users was rolled back. The pricing page now lists Claude Code as included with Pro, Max, Team and Enterprise.
You need one of five accounts: Pro, Max 5x, Max 20x, Team, or an Anthropic Console account with an API key. Each one unlocks Claude Code in the terminal, the desktop app, the VS Code extension, the JetBrains plugin, and the iOS app. The same login works across all surfaces, so you only pay once.
Pro at $20 a month is the entry point. You get Claude Code on Sonnet, weekly usage caps that Anthropic tunes against the average Pro developer, and access to the desktop, terminal and IDE clients. Max 5x at $100 a month bumps the cap roughly five times higher and unlocks Opus access. Max 20x at $200 a month is twenty times the Pro cap and is the only plan that supports the highest effort tier, useful if you're doing long agentic runs across a big codebase. Team and Enterprise are seat-based for orgs that want SSO and central billing.
The API key path is different. You don't pay a flat fee. You add a credit card to the Anthropic Console, generate a key, point Claude Code at it with ANTHROPIC_API_KEY, and you get charged per token. Cheap if you barely use it, brutal if you run agents all day.
Pro is $20 a month, or $17 a month billed annually. Max 5x is $100 a month. Max 20x is $200 a month. Team is $25 per seat per month, or $20 per seat on annual billing. Enterprise is $20 per seat plus usage. The API path has no monthly fee but you pay per million tokens, with Sonnet 4.6 priced lower than Opus 4.7.
Anthropic publishes the full breakdown on the official pricing page, but the page hides the most important detail behind a hover. The Pro plan is rate-limited weekly, not just per-session. If you push hard for two days, you can burn through your weekly cap and get throttled until the cycle resets. That's the real cost on Pro for power users, and it's the reason most people I know on Pro upgrade to Max within a couple of months.
Yes, briefly. Between April 21 and April 22, 2026, Anthropic updated the Pro plan feature list to show an X next to Claude Code instead of a checkmark. New users signing up to Pro saw Claude Code disabled. Developers spotted it within hours and the forums lit up. Anthropic confirmed it was a test affecting roughly 2% of new users, said existing subscribers were not affected, and reverted it shortly after.
Anthropic's growth lead said at the time: "When we do land on something, if it affects existing subscribers you'll get plenty of notice before anything changes." The full Register report is worth reading for context. The takeaway: the official pricing page is the source of truth on any given day, and right now Claude Code is included with Pro again.
Yes. Pro at $20 a month includes Claude Code on Sonnet 4.6 across the terminal, desktop, IDE and web. You log into Claude Code with the same email you use for claude.ai and your subscription is detected automatically. No extra step, no separate billing, no API key needed.
Pro is genuinely useful for solo developers, side projects and content work. I built three internal tools and shipped two client projects on Pro before I hit the weekly cap. Where Pro starts to hurt is if you're running long agent loops, refactoring across hundreds of files, or you sit in plan mode for hours generating large outputs. The cap arrives faster than you expect because agentic work consumes far more context than chat.
Not in the way most tools mean it. There is no free tier for individual developers. Students don't get a discount on Claude Code specifically, though Anthropic occasionally runs Claude for Education promotions on the chat product. Open source maintainers, hackathon entrants and indie hackers all pay the same $20 a month.
The closest thing to free is the Anthropic API trial credit, which Anthropic hands out occasionally when you create a Console account. That credit, if you get it, runs through the same key you'd use with Claude Code. So technically you can run Claude Code for free until the credit runs out, but it's small and it doesn't refresh. Treat it as a tasting menu, not a free tier.
Only if you tell it to. By default, Claude Code authenticates to your Pro or Max subscription, so usage counts against your subscription's weekly cap, not your API balance. If you set the ANTHROPIC_API_KEY environment variable, Claude Code switches to API billing and every prompt, file read and tool call eats your token credit.
This is the trap most teams fall into. They set the API key for one experiment, forget to unset it, and a week later their Console balance is half drained. If you're on Pro or Max, do not export ANTHROPIC_API_KEY in your shell profile. Use it only when you genuinely need API billing, like agent SDK work or running Claude Code in CI through GitHub Actions.
For most people, Pro at $17 a month on annual billing. That's the floor. If you use it for under an hour a day on side projects, Pro is plenty and the $204 a year is cheaper than almost any developer tool you'll buy.
For heavy users, Max 5x at $100 a month is cheaper than spinning up the same workload on the API. The math: a single hour of agent time on Opus 4.7 with serious context can cost $5 to $15 in API tokens. Twenty hours of that a month and you're already at Max territory. The flat fee buys predictability. For occasional users with bursty workloads, the API key route can be cheapest, especially if you only use Claude Code once a week. Set a budget alert in the Console and it stays under control.
ChatGPT has a free chat tier. Claude Code does not. That's the core difference. OpenAI's free ChatGPT plan lets you talk to GPT-5 with limits, no card on file. Anthropic's free Claude.ai plan lets you talk to Claude Sonnet, also no card. Neither free plan gives you a coding agent that runs in your terminal.
If you want a comparable free agentic coding experience, the closest options are GitHub Copilot's free tier for verified students, OpenAI's Codex CLI through a paid ChatGPT Plus or Pro subscription, or open-source agents you run against a local model. Each has trade-offs and none of them match Claude Code's multi-surface UX. I covered the gap between Claude Code and the rest of the field in Claude Code just got 10x more powerful if you want the feature-by-feature breakdown.
I see this line on Twitter and LinkedIn most weeks: "Claude Code is basically free if you already pay for Claude Pro." That's true today. It's also fragile. The April 2026 test proved Anthropic can pull the bundle on a whim, and the next time it might not be a 2% experiment. If you're building a business around Claude Code being included with Pro, you're building on a feature that's bundled at Anthropic's discretion.
The honest framing: Claude Code is included with Pro right now. Treat it as a $20 a month standalone tool that happens to bundle a chatbot, not a free add-on to your chatbot subscription. Price your work that way, and you don't get caught out if the bundling changes.
Yes for anyone shipping code or building automations weekly. No for anyone who hasn't actually written a project with it yet. Pro at $20 buys you a real engineering teammate that reads your codebase, edits files, runs commands and ships work in the background. If you use it for an hour a day for a month, the cost-per-hour is under a dollar.
If you're a beginner, start on Pro for one month. Don't jump to Max until you hit the weekly cap twice. Most people never hit it. If you're a heavy agent user running long autonomous tasks across multiple repos, go straight to Max 5x. You'll save money over the API route and you'll stop worrying about token budgets every prompt. If you're testing Claude Code for a specific project and won't touch it again for weeks, the API key route is fine. Set a $20 budget cap in the Console and forget about it.
Not as a published feature. Anthropic occasionally seeds new Console accounts with a small API trial credit you can spend on Claude Code through an API key, but it's not advertised, it doesn't refresh, and it's small. The closest thing to a trial is the Pro plan's monthly billing. Cancel inside 30 days and you've spent $20 to take Claude Code for a real test drive.
It's included, not free. Max 5x at $100 and Max 20x at $200 a month both bundle Claude Code along with higher rate caps, Opus 4.7 access and the highest effort tier on Max 20x. You're paying for the model and the cap, and Claude Code rides along.
The VS Code extension itself is free to install from the Marketplace. The agent inside it is not. The extension authenticates to the same Pro, Max, Team or API account you use everywhere else, and usage counts against that plan. There's no separate VS Code-only tier.
If you're on Pro, Max, Team or Enterprise, usage is metered against your plan's weekly cap and you don't see line-item costs. If you're on the API path, every prompt and tool call is billed per token to the credit card on your Console account. You can track real-time spend with the /usage command inside Claude Code.
Only on the API path. Subscription users (Pro, Max, Team, Enterprise) pay a flat monthly fee and run against a weekly usage cap, not per-request billing. Once you exceed the cap, you're throttled until the next weekly window resets. API users pay per million tokens of input and output, billed continuously.
Yes. Sign up for Pro, log into Claude Code with claude /login, and unset ANTHROPIC_API_KEY from your shell. Claude Code will route through your subscription instead of your API key. The reverse also works. Start the month on Pro, hit the cap, export the API key for the rest of the month, and Claude Code switches over instantly.
Not as a dedicated student plan. Anthropic has run Claude for Education promotions on the chat product, but Claude Code itself doesn't have a student discount published. The cheapest student route is the same as anyone else: Pro at $17 a month annual, or the API path with a tight Console budget cap.
If this guide was useful, these go deeper on adjacent topics:
Primary sources referenced in this guide:
Pricing is the easy part. The hard part is figuring out what to build with it. The free Claude Code Blueprint is a 60-minute walkthrough that takes you from install to your first shipped automation, no coding required. If you want a structured 30 days with weekly live calls, the Claude Code Challenge runs in cohorts. And if you're curious what real students have built with this, Kevin's case study is the one I send to anyone asking whether $20 a month is worth it.
No. Claude Code requires a paid plan. Anthropic's setup documentation says it plainly: Claude Code requires a Pro, Max, Team, Enterprise, or Console account, and the free Claude.ai plan does not include access. The cheapest legitimate route is Pro at $20 a month ($17 annual). The other path is a pay-as-you-go API key billed per token through the Console.
One of five accounts. Pro at $20 a month is the entry point with Claude Code on Sonnet. Max 5x at $100 bumps the cap roughly 5x and unlocks Opus. Max 20x at $200 is 20x the Pro cap and unlocks the highest effort tier. Team is seat-based for orgs. Or use an Anthropic Console account with an API key for pay-per-token billing.
Only if you tell it to. By default Claude Code authenticates against your Pro or Max subscription, so usage counts against your plan's weekly cap. If you set ANTHROPIC_API_KEY in your environment, Claude Code switches to API billing and every prompt and tool call eats your token credit. This is the trap most teams fall into. Do not export the key in your shell profile if you are on a subscription.
For most people, Pro at $17 a month on annual billing. That is the floor. For heavy users running long agent loops on Opus, Max 5x at $100 a month is cheaper than the same workload on the API (one hour of Opus agent time can cost $5 to $15 in tokens). For occasional bursty use once a week, the API key with a $20 budget cap in the Console is cheapest.
Not as a published feature. Anthropic occasionally seeds new Console accounts with a small API trial credit you can spend on Claude Code through an API key, but it is not advertised, does not refresh, and is small. The closest thing to a trial is Pro's monthly billing. Cancel inside 30 days and you have spent $20 to take Claude Code for a real test drive.
Briefly. Between April 21 and April 22, 2026, Anthropic ran a test that pulled Claude Code from Pro for roughly 2% of new users. Existing subscribers were not affected. Anthropic reverted the test within hours after backlash on the forums. The Pro plan now includes Claude Code again. The official pricing page is the source of truth on any given day.
No dedicated discount. Anthropic has run Claude for Education promotions on the chat product, but Claude Code itself does not have a published student rate. Open source maintainers, hackathon entrants, and indie hackers all pay the same $20 a month. The cheapest student route is Pro on annual billing at $17 a month, same as anyone else.
Five interactive lessons. Install Claude Code, build your first automation, and deploy it live on the internet — all in under an hour. Free, no coding required.
Grab the Blueprint →