Claude Code Pricing: Plans, API Billing, Team Costs

Claude Code costs nothing to install, but it needs a paid way to run. The cheapest is a Pro plan at $17 a month billed annually ($20 monthly). Heavy users move to Max ($100) or Ultra ($200), or pay per token through a Console account. Anthropic's enterprise average is about $13 per developer per active day. Prices checked 2026-10-31.

Prices checked 2026-10-31

Claude Code has no price of its own. You pay through one of three meters.

The plan prices come from claude.com/pricing and the model rates from Anthropic's API docs. For the full plan comparison, see Claude pricing. For tokens alone, see Claude API pricing. For the tool itself, see Claude Code.

Way one: a Claude subscription

Pro, Max, Ultra, Team and Enterprise all include Claude Code. Free does not. Anthropic describes it as "one unified subscription" for the web, desktop and mobile apps and the terminal.

You log in with the same credentials you use for Claude. The IDE extensions for VS Code, Cursor and other VS Code forks, and JetBrains use the same login and count toward the same limits.

The shared pool

All Claude activity draws from one allowance: claude.ai chat, the desktop app, mobile and Claude Code. A morning of chat reduces what is left for the afternoon's coding. This is the surprise developers hit first, and it is spelled out in Anthropic's help center.

What a limit looks like

Paid plans have a five-hour session limit and a weekly limit. The five-hour window starts at your first message. The weekly limit resets at a fixed time assigned to your account. Claude Code warns you as you approach either.

If you hit the session limit mid-response, Claude Code may keep going for a short, capped wrap-up so it can finish the step in progress. You will see "Usage limit reached · wrapping up". Per Anthropic's terms, that allowance counts toward your weekly limit, applies on Pro at most once per weekly period, and can apply at each five-hour limit on Max. It needs Claude Code 2.1.277 or later, does not apply to API keys or cloud providers, and is rolling out gradually. It is not a credit and has no cash value.

When you are properly out, the choices are:

1. Wait for the reset.
2. Upgrade (Pro to Max, Max to Ultra). New limits apply at once.
3. Turn on usage credits.
4. Switch to a Console account for a heavy sprint.

Usage credits in Claude Code

Usage credits turn a hard stop into pay-as-you-go at standard API rates. Run /usage-credits after logging in through /login with your claude.ai account. The command does not exist under API-key authentication. You prepay a balance, can set a monthly spend limit, and can turn on auto-reload. When you reach the cap, Claude Code offers to raise or remove it without leaving the CLI.

Two details are worth noting. Usage credits are billed apart from the subscription and appear as extra charges. And the monthly API credits that come with Max, Ultra and Team cannot be spent on Claude Code at all; they only work on the Claude Platform.

Fable and Claude Code

Neither Fable model is the default on any plan. You pick one explicitly with /model. On Pro and Team Standard seats it bills to usage credits from the first token, and the picker shows "Requires usage credits" on that row. On Max, Ultra and Team Premium seats it is included up to 50 percent of your weekly limit. Fable 5.1 needs Claude Code 2.1.255 or later.

Way two: the Console API

With a Console account, Claude Code bills per token at API rates. There is no five-hour limit, only your tier's rate limits and spend cap. The first time you authenticate with Console credentials, Claude Code creates a workspace called "Claude Code" for cost tracking. You cannot create API keys in it, and its traffic counts toward your organization's rate limits unless you set a workspace limit.

Claude Code works out a dollar figure locally from token counts at list price. The /usage session block shows it. Anthropic notes it is an estimate, and the Console usage page is authoritative. For subscribers that figure is not what you are billed.

Model rates on the API today:

Claude Code leans hard on prompt caching, because every request re-sends the whole conversation. Anthropic says it re-reads that history at the cached rate. A cache miss, such as the first message after a break longer than the cache lifetime, reprocesses the full context at the full rate.

Way three: a cloud provider

Claude Code runs through Amazon Bedrock, Google Cloud's Agent Platform, Microsoft Foundry and Claude Platform on AWS. The cloud provider invoices you, and the aliases map to different models. On the Anthropic API, sonnet is Sonnet 5.5 and opus is Opus 5.5. On Bedrock and Google Cloud the sonnet alias is Sonnet 4.5, and on Foundry opus is Opus 4.6. Pin a model with a full ID or ANTHROPIC_DEFAULT_SONNET_MODEL if that matters. Sonnet 5.5 needs Claude Code 2.1.284 or later and Opus 5.5 needs 2.1.280.

What it costs per developer

Anthropic's costs doc gives the one cost figure it stands behind: across enterprise deployments, the average is around $13 per developer per active day and $150-250 per developer per month, with costs below $30 per active day for 90 percent of users.

That is an average of API-billed usage across many companies. It is not a promise, and it will mislead you in both directions. A developer on Sonnet with tight context habits can sit well below it. A team running multiple instances or automation can sit well above. Anthropic's own advice is to start with a small pilot group and measure before wider rollout.

The useful comparison is with seat prices.

For a developer who would cost $150-250 on API billing, a $100 Max plan or a $100 Premium seat is a lot cheaper, provided the 5x allowance covers their use. The subscription caps your spend and the API does not. Usage credits sit between the two.

Team and Enterprise controls

On Team and Enterprise plans each member draws from a per-seat allowance, on the same rolling five-hour and weekly windows, shared with chat and Cowork. Size depends on seat tier. The controls are in the claude.ai admin console, not the Claude Console.

• See spend. Org analytics shows estimated spend per user and per model, with CSV export, updated daily. It covers usage-credit spend only; use inside the seat allowance is not metered in dollars.
• See adoption. An analytics dashboard shows daily active users, sessions and contribution metrics.
• Cap spend. The seat allowance is the default ceiling. Turn on usage credits and set limits at organization, group or individual level to let people continue past it.
• Per-user data. Enterprise has an analytics API. API customers use the Console dashboard and the Claude Code Analytics API with an Admin key.

For API organizations, spend is capped with workspace spend limits and workspace rate limits. Anthropic's doc includes per-user rate limit recommendations by organization size.

Admins should know which message a developer is seeing. "You've hit your session limit" or "weekly limit" is a seat window, shared across all models, so switching models in /model does not restore access. "You've hit your individual spend limit" means usage credits hit a cap an admin set. A context or auto-compact warning is neither; it just means a long conversation.

Why your usage climbs

A session left open for hours can burn far more allowance than your activity suggests. Anthropic's docs list the reasons.

• Long context. The whole conversation is sent with each request.
• Cache misses. The first message after a long pause reprocesses everything.
• Scheduled tasks and goal check-ins. They fire even when you are idle, each carrying the full context.
• Subagents and agent teams. Each one sends its own requests, and each active teammate keeps consuming until it exits.
• Compaction. /compact reads the whole conversation, so it is itself a large request.

Run /usage on a plan. It shows an attribution breakdown by skill, subagent, plugin and MCP server, and recommends changes for anything behind 10 percent or more of your recent usage.

How to spend less

These are straight from Anthropic's costs doc.

See also our guides on Claude Code best practices and MCP servers for what to keep connected.

Who should pick which

Our take

The question people type is "how much does Claude Code cost", and the honest answer has a floor and no ceiling. The floor is $17 a month. The ceiling depends on whether you pay by plan or by token.

From the documents, a plan is the safer default for individuals, because the cost is capped and you find out about limits by hitting them rather than by getting an invoice. API billing wins when the usage is automated, bursty or tied to a project budget. Usage credits are the bridge, and the monthly spend cap is the setting to touch first.

One trap deserves repeating. If ANTHROPIC_API_KEY is set in your environment, Claude Code uses that key rather than your subscription, and you are billed API rates. Anthropic's help center calls this out. If a Max subscriber's bill looks wrong, check that variable before anything else.

How Claude Code compares with other tools on price is on AI coding agent pricing compared, with Cursor pricing, GitHub Copilot pricing and Codex pricing for the specifics. We have not measured any of them against each other.