Claude Code Pricing Explained: What It Costs per Seat, per Team, and via API (2026)
Von Manuel Gick·Certified AI & Automation Expert
Claude Code pricing works in two ways: it is included in every paid Claude subscription (Pro from $17 a month, Max from $100, Team seats from $20), and it can run on pay-as-you-go API credits billed per token. Which one is cheaper depends on how many hours a day someone actually codes with it.
This article explains what Claude Code costs per seat, per team, and via the API, when the API beats a subscription, and how the price compares with Codex, Cursor, and OpenCode – with prices as listed by the vendors on 5 September 2026.
Claude Code is Anthropic's coding agent; it runs in the terminal, your editor, or the desktop app. We use it to build client software – what that looks like day to day is described in Agentic coding with Claude Code, and what we build with it is on our custom software page. This article is only about the money.
How much does Claude Code cost per seat?
Claude Code is included in all paid Claude plans and draws from the same usage pool as chatting with Claude on the web or in the apps. The plans for a single user, as listed on 5 September 2026:
| Plan | Price | Claude Code included? | Usage |
|---|---|---|---|
| Free | $0 | No | Chat only |
| Pro | $17 per month billed annually ($200 up front), $20 billed monthly | Yes | Base usage per 5-hour session, plus weekly limits |
| Max | From $100 per month | Yes | 5x or 20x the Pro usage, in two tiers (see pricing page) |
All prices are US dollars as listed by Anthropic on 5 September 2026, before VAT – for a German company, add the tax and the exchange rate of the day.
Two things matter more than the sticker price. First, usage is measured per 5-hour session with a weekly cap on top, and everything you do across web, desktop, mobile, and Claude Code draws from the same pool. Second, when you reach a limit, you either wait for the next window or switch to pay-as-you-go API credits through a Console account – the choice is yours, the billing changes, the tool stays the same.
Prices change; always check the official pricing page before you budget.
What does Claude Code cost for a team?
For companies, Anthropic offers Team and Enterprise plans. Team is for 2 to 150 people and mixes two seat types:
- Standard seat: $20 per seat per month billed annually ($25 monthly) – all Claude features, more usage than Pro.
- Premium seat: $100 per seat per month billed annually ($125 monthly) – 5x more usage than a standard seat, designed for people who code with Claude Code all day.
You can mix seat types, so a company can give two developers premium seats and the rest of the office standard ones. According to the pricing page, Team plans come with central billing and administration, single sign-on, and no model training on your content by default.
Enterprise costs $20 per seat per month plus usage billed at API rates, annual billing only; the pricing page lists admin-set spend limits per user and organization and role-based access among its features. In other words: Enterprise turns the flat subscription into a base fee plus consumption – the right shape for organizations that want central billing and pay for actual usage afterwards. Which admin and compliance features come with each plan changes regularly; check the current page before signing.
When is the API cheaper than a subscription?
Claude Code can run entirely on API credits instead of a subscription. Then you pay per token – the units of text the model reads and writes. The list prices per million tokens (input / output) as of 5 September 2026:
| Model | Input | Output | Typical use in Claude Code |
|---|---|---|---|
| Fable 5.1 | $10 | $50 | Long-running, complex agent work |
| Opus 5 | $5 | $25 | Complex coding and architecture |
| Sonnet 5 | $2 | $10 | Everyday coding and agents |
| Haiku 4.5 | $1 | $5 | Fast, cheap subtasks |
Prompt caching means repeated context isn't billed at full price every time, and batch processing is 50% cheaper – both matter for agents that run unattended.
What does that mean per developer? The only public figure that isn't a guess comes from Anthropic's own documentation: across enterprise deployments, the average cost is around $13 per developer per active day and $150–250 per developer per month, with 90% of users staying below $30 per active day. Set against a Max plan at $100 or a premium team seat at $100–125, the rule of thumb is:
- Light or occasional use (a few sessions a week): the API is usually cheaper – you pay only for what you use.
- Daily, heavy use (a developer who works with the agent most of the day): a Max plan or a premium seat is usually cheaper and, above all, predictable.
- Unattended agents and automation (Claude Code running in pipelines, overnight jobs, multiple parallel instances): the API, with caching and batch processing, because subscriptions are built for a person at a keyboard, not for a fleet. And a fleet gets expensive in a second place: every system an agent talks to needs a connection – usually built as an MCP server, explained in What is MCP?.
How we do it ourselves: our own AI agents – more than 40 of them – run in Langdock, not in Claude Code; Claude Code is what we build client software with. Both budgets sit in different places, and that is exactly the split most companies end up with: a chat platform for the office, a coding agent for the developers. To give you one honest number of our own: our bookkeeping runs around 90% automated – roughly 60 hours a month before, roughly 15 today – built with exactly this kind of agent work.
Claude Code vs. Codex vs. Cursor vs. OpenCode: how the pricing models differ
The four tools most businesses compare are priced on four different logics. The table shows the structure, not a winner – which one is cheapest depends on your usage pattern. All prices as listed on the vendor pages on 5 September 2026.
| Tool | Pricing logic | Entry price | Team pricing |
|---|---|---|---|
| Claude Code (Anthropic) | Included in Claude subscriptions, or API per token | Pro $17–20/month; Max from $100 | Team seats $20 / $100 (premium); Enterprise $20 + API usage |
| Codex (OpenAI) | Included in ChatGPT plans, usage limits per plan | Free $0; Go $8; Plus $20; Pro from $100 (5x/20x) | Business $20/user/month annual ($25 monthly) |
| Cursor | Editor subscription with an included usage amount, then usage-based billing | Hobby free; Individual Pro $20/month; Pro+ and Ultra above | Teams $40/user/month |
| OpenCode | Open source, free; you bring your own model from any provider or use its own model access, billed per token | $0 for the tool | None – costs are the model tokens you use |
How to read that:
- Claude Code and Codex are priced the same way: the agent comes with a chat subscription you may already be paying for. If your team already has ChatGPT Business or Claude Team, the coding agent is a feature, not a new line item. The difference is the model behind it and how much usage each plan includes.
- Cursor is an editor first – a modified version of the widely used VS Code. Its subscription includes a usage amount, and heavy agent use tips into usage-based billing. For a team that lives in one editor, that's convenient; for unattended agent runs, it's the wrong shape.
- OpenCode removes the tool cost entirely and moves everything to model tokens. That's the cheapest option if someone on your team wants to manage API keys and model choice – and the most technical one. You can also run Claude models through OpenCode's own model access, billed per token; on 5 September 2026 its rate for Opus 5 matched Anthropic's list price ($5/$25 per million tokens), but check the current rates there – they are not automatically identical. So "OpenCode vs. Claude Code" is usually a question of interface and control, not of the model you get.
Does Claude Code run on Windows – and does that change the price?
Yes. Claude Code runs natively on Windows as well as macOS and Linux, and the price is the same on every platform – it's tied to your Claude plan or API account, not to the operating system. If your developers work on Windows machines, there is no extra license and no separate product.
Claude Code best practices that keep the bill down
Whatever plan you choose, the same habits decide whether a developer lands at the bottom or the top of that $150–250 range – and whether an unattended agent run surprises you at the end of the month:
- Pick the model per task. Sonnet 5 for everyday work at $2/$10 per million tokens, Opus 5 or Fable 5.1 only when the task really needs it.
- Keep sessions focused. Long sessions accumulate context, and every message re-reads it. Clear the session when you switch tasks.
- Keep the cache warm. Prompt caching makes repeated context cheap; according to Anthropic's cost documentation, changing tool definitions mid-session is one of the things that invalidates it.
- Set spend limits. On Enterprise, admins can cap spend per user and per organization; on the API, Claude Code has a command-line option for a maximum budget per session, and the Console shows usage. Do it before the first surprise, not after.
- Measure per developer, not per company. The docs' benchmark – around $13 per active day on average – is only useful if you know your own number. Start with a small group, read the usage page after two weeks, then decide between subscription and API.
Frequently Asked Questions
Is Claude Code free?
Not on its own. The Free Claude plan doesn't include Claude Code according to the feature table on Anthropic's pricing page (5 September 2026); the cheapest way in is the Pro plan at $17 per month billed annually or $20 monthly, which includes Claude Code alongside chat. Alternatively, you can run Claude Code on pay-as-you-go API credits with no subscription at all – then you pay only for the tokens you use, which is usually cheaper for occasional use.
How much should I budget for Claude Code per developer?
Budget one number, not a plan: what does an active coding day cost this person? Anthropic's documentation puts the enterprise average at around $13 per active day and $150–250 per month, with 90% of users under $30 a day. If your developer is above that, a Max plan or premium seat caps the cost at $100–125; if well below, Pro or the API is cheaper. Read your own usage page after two weeks – that beats any estimate.
Is Claude Code cheaper than Codex or Cursor?
On the $20 tier, Claude Pro, ChatGPT Plus, and Cursor Individual Pro sit side by side; below that, Codex and Cursor offer free and cheaper entry levels, while Claude Code needs at least Pro. The real difference is the usage included and the model behind it. For a team, compare Claude Team seats ($20 standard, $100 premium) with ChatGPT Business ($20 per user) and Cursor Teams ($40 per user).
Can my whole company use one Claude Code subscription?
No. Individual plans are personal, and Team plans are billed per seat, 2 to 150 users, with seat types you can mix – that is the structure Anthropic sells to companies. A shared login also exhausts the usage limits within a day, so it doesn't work in practice either. The right setup for a company is a Team plan with mixed seats: premium seats for the people who code with the agent all day, standard seats for everyone else, or an Enterprise plan with base fee plus API usage.
What happens when I hit the usage limit?
Usage is measured per 5-hour session with weekly caps on top. When you reach a limit, you wait for the next window or switch to pay-as-you-go API credits through a Console account – the tool stays the same, only the billing changes. On Enterprise, admins can set spend limits so that a single heavy user can't run up an unexpected bill; on the API, a per-session budget option does the same for automated runs.
The next step
If you're weighing Claude Code for your own team, the pricing is the easy part; the harder question is which of your processes are worth building software for at all – and whether custom software or an off-the-shelf tool is the right answer. That's what a free Time Potential Analysis answers in 45 minutes: no sales call, and you leave with your three biggest time drains in writing – and a clear view of where custom software or AI agents would pay off. Book your Time Potential Analysis →
