Claude Code enterprise cost: seats vs API

What does Claude Code cost for an enterprise?

Claude Code enterprise cost takes one of two shapes, and picking the wrong one is where a budget goes wrong. On a Team plan the seat price includes the usage: you pay a fixed amount per person each month, and Claude Code draws on an allowance attached to that seat. On a current Enterprise plan the seat is cheap and carries no usage at all, so every token your developers consume is billed on top at standard API (application programming interface) rates.

The figures below are list prices published on claude.com/pricing, read on 31 October 2026. They are US dollars per user per month on annual billing. Prices move, so open the page yourself before you sign anything.

[
  {
    "plan": "Team standard",
    "usd_per_user_month": 20
  },
  {
    "plan": "Team premium",
    "usd_per_user_month": 100
  },
  {
    "plan": "Enterprise seat",
    "usd_per_user_month": 20
  }
]

A Team standard seat is $20 per user per month on annual billing, or $25 if you pay monthly. A Team premium seat is $100 on annual billing, or $125 monthly. The Enterprise seat is $20 per user per month, billed annually, and that number is only the entry fee. This guide stays on Claude Code. For how the plans differ on everything else, read the Team and Enterprise plan comparison.

What one seat includes

Claude Code is included with every Team plan seat, standard or premium. On new and self-serve Enterprise plans it is included with the single Enterprise seat, so there is nothing extra to buy. HIPAA-ready Enterprise organizations are the exception: Anthropic provisions them with separate Chat and Chat plus Claude Code seat types, and they are not eligible for the single Enterprise seat billing model, so check which seat type each developer holds. The same seat covers Claude Code in the terminal and in supported editors, including VS Code, Cursor and other VS Code forks, and JetBrains editors such as IntelliJ and PyCharm. Editor use counts against the same allowance as the terminal, so a developer cannot get a second allowance by switching window.

On Team and seat-based Enterprise plans, each member's usage draws from a per-seat allowance that resets on a rolling five hour window and on a weekly window. That allowance is shared with Claude chat and Cowork. This matters in practice, because a person who spends the morning in chat has less left for Claude Code in the afternoon. Anthropic's help centre states that a Team premium seat includes 6.25 times the Pro plan's per-session usage allowance. A multiplier against Pro is what you are buying, not a token bucket you can read off a page. Anthropic prices its individual Max and Ultra plans on the same logic, with 5x and 10x Pro usage per week respectively, so the Max and Ultra plans are a useful reference for what a usage multiplier actually feels like day to day. Verified Ultra accounts have no five-hour limit; a temporary limit applies before verification.

Usage inside a seat allowance still has an estimated dollar cost. Claude Code exports claude_code.cost.usage through OpenTelemetry in USD, incremented after each API request, with user attributes for per-developer reporting, including subscription usage. These estimates do not represent additional subscription charges. Usage credits create separately billed spend, which is covered below.

Team plans also include monthly API credits, pooled across seats and capped at $500, which an Owner or Primary Owner claims into a linked Claude Console organization. Those credits do not cover interactive Claude Code or extra usage after the seat allowance runs out.

Standard seats and premium seats on Team

A Team plan needs at least 2 members and supports up to 150 seats. Only Owners and Primary Owners can buy seats. You buy a pool of seats of each type in Organization settings, then assign people into them.

Moving a person between Standard and Premium does not require a purchase if you already own a free seat of that type. Changing the Tier dropdown on a member is enough. You can also park someone on "No seat assigned" to free their seat for somebody else. Reassignment is meant for normal team changes, such as someone joining, leaving or changing roles, and Anthropic says plainly: "It isn't designed for rotating one seat among several active users." Each seat belongs to one named person, so every developer doing heavy work needs a premium seat of their own.

Two billing details decide whether over-buying hurts. New seats and upgrades are prorated and charged immediately, so adding capacity mid-cycle is cheap. Going the other way is slow: moving a member from Premium down to Standard takes effect at your next renewal, and removing a member does not refund anything. You have to reduce the total seat allocation to lower the bill, and even then the credit is not immediate. Buy the seats you need now and add more later.

The Enterprise seat is a thin fee plus metered usage

Enterprise works the other way round. The billing help centre is explicit that the seat fee "doesn't include any usage" and that "Every token your team consumes is billed separately at standard API rates." The bill therefore has two lines: a fixed seat fee, and consumption. One exception changes the rate rather than the shape: if your organization turns on US-only inference, usage of Claude Opus 4.6, Claude Sonnet 4.6 and later models is billed at 1.1x standard API rates. A self-serve Enterprise plan needs a minimum of 20 seats and a sales-assisted one needs a minimum of 50. On both, seats are priced per user per month and billed annually.

That is why the Enterprise seat costs less than a Team premium seat while including the same Claude Code access. You are not paying for an allowance, because there is no allowance. A quiet Enterprise seat is genuinely cheaper than a Team premium seat, and a busy one is not. For the rest of what the Enterprise contract carries, see Enterprise plan pricing.

Older Enterprise agreements had more seat types, including Chat, Chat plus Claude Code, Standard and Premium. Those move to the single Enterprise seat at renewal, except in HIPAA-ready organizations, which Anthropic provisions with separate Chat and Chat plus Claude Code seats and says "are not eligible for the single Enterprise seat billing model." If your organization is on that seat model and a developer reports that Claude Code will not authenticate, check their seat type first. A plain Chat seat has no Claude Code access until an owner reassigns it in Organization settings.

Self-serve or sales-assisted, and why it changes your invoice

Enterprise comes in two buying paths, and they meter the same usage differently. A self-serve organization buys usage credits upfront and draws them down. A sales-assisted organization is billed monthly in arrears on actual consumption. Enterprise plans purchased through AWS Marketplace are also billed monthly in arrears, with payment through AWS. The paths also start from different floors: self-serve needs at least 20 seats, and a plan bought through Sales needs at least 50. Owners can add seats at any time during the annual term. Taking seats away is where the paths split: on a self-serve plan a reduction only takes effect at renewal, and on a sales-assisted plan it goes through your account manager. Plan the first seat count around that before you onboard a new team, because shrinking it later is slower than growing it.

Credits are shared across the whole organization, so any member can draw from the same pool. Pooling is convenient and it is also the failure mode. One developer running an unattended loop overnight spends from the same pool as everyone else, and the spend report only shows it after the next daily refresh; OpenTelemetry export provides cost estimates in near real time. Set per-member spend limits before you turn credits on, not after.

What an admin can actually control

The controls live in the claude.ai admin console, not in the Claude Console used for API keys. Which set applies to you depends on how your developers signed in.

• Seat purchase, assignment and tier changes, in Organization settings
• Usage credits on or off, which decides whether anyone can work past the seat allowance
• Spend limits for the whole organization and for individual members on both Team and Enterprise. Group spend limits are Enterprise only: Anthropic says "Groups and group spend limits are available for Enterprise plan organizations"
• Managed policy settings pushed to every Claude Code user, covering tool permissions and file access
• Single sign on (SSO), just-in-time provisioning, domain capture and role-based permissioning on both Team and Enterprise. The claude.com/pricing comparison table still marks domain capture and role-based access as Enterprise only, but Anthropic's Team plan help article lists all four for every Team seat

The seat allowance is itself the default ceiling. If you never enable usage credits, a member who exhausts their window simply waits for it to reset and sees the reset time in the message. That is a hard cap with zero finance risk and a real productivity cost. Enabling credits with per-member limits converts that hard stop into a controlled overspend, which is usually what a working team wants.

Where you see what a developer spent

The spend report in org analytics gives estimated spend per user and per model, with a comma separated values (CSV) export. The columns are the user's email, account UUID, product, model and model family, request count, prompt tokens, completion tokens, net spend and gross spend. It refreshes daily and runs one day behind, so it is a review tool, not an alarm. It is visible to Team Owners and Primary Owners, and to Enterprise Owners and Primary Owners. Enterprise Admins can see the rest of analytics but not spend.

One limit catches people out. On Team and seat-based Enterprise plans, the spend report covers usage-credit spend and only appears once usage credits are turned on; estimates for usage inside the seat allowance are available separately through OpenTelemetry. If your Team plan has credits switched off, the report is empty by design and nothing is broken. A current usage-based Enterprise plan has no allowance, so its export covers the organization's full usage.

Adoption is measured separately from spend. The analytics dashboard at claude.ai/analytics/claude-code shows daily active users, sessions, lines of code accepted and suggestion accept rate, and, once the GitHub app is connected, merged pull requests and lines of code shipped with Claude Code assistance. Its Value tab also shows cost per commit, pull request and session; per-user cost estimates come from the spend report or from OpenTelemetry export. On Enterprise, a Primary Owner can create a key with the read:analytics scope and pull per-user usage and cost numbers through the Enterprise Analytics API. Team plans get the CSV export instead.

A developer can check their own position without asking anyone. Running /usage inside Claude Code shows the plan usage bars, and d or w switches between the last 24 hours and the last 7 days.

Total cost:            $0.55
Total duration (API):  6m 20s
Total duration (wall): 6h 33m 10s
Total code changes:    0 lines added, 0 lines removed
Usage by model:
   claude-sonnet-4-6:  1.2k input, 5.3k output, 940.0k cache read, 50.0k cache write ($0.55)

Read that dollar figure carefully. It is computed locally from token counts at list price, so it ignores contracted discounts unless an administrator has deployed a modelPricing table in managed settings, in which case the line notes that it uses your organization's configured rates. Either way it is an estimate aimed at API users. On a seat it does not correspond to a line on your invoice. It also only covers sessions on that one machine. For an inventory of what each reporting surface actually answers, see the Claude Code spend tracking tools.

Seats or API keys: the decision for ten developers

This is the real question, and Anthropic has published enough to answer it. On 19 August 2026 the Claude Enterprise consumption guide gave per-user monthly starting points for budgeting. When we reopened the guide on 27 October 2026 those figures were no longer on it, so read the chart below as Anthropic's August planning estimates for typical deployments, not a current quote.

[
  {
    "label": "Power user, top 10%",
    "claude_code_usd": 500,
    "cowork_usd": 100,
    "chat_usd": 90
  },
  {
    "label": "Typical user, mean",
    "claude_code_usd": 215,
    "cowork_usd": 40,
    "chat_usd": 30
  },
  {
    "label": "Light user, median",
    "claude_code_usd": 40,
    "cowork_usd": 10,
    "chat_usd": 5
  }
]

Work it through for ten developers. Ten Team premium seats cost a fixed $1,000 a month. On self-serve Enterprise, ten developers still pay for the 20 seat minimum, so the seat line is at least $400 a month, billed annually, however few people use it. Buying through Sales raises that floor to 50 seats. Ten metered users at the mean Claude Code figure of $215 cost $2,150 in usage, plus $400 in self-serve Enterprise seat fees, $2,550 in total. On those numbers the premium seat wins by more than half. Now run it for a light team at the median of $40: $400 in usage plus $400 in seats is $800, so the result flips: the premium seats are now the expensive choice, by $200 a month. A single power user at $500 a month costs five times what their seat does.

The distribution is what decides it, not the average. The Claude Code documentation reports roughly $13 per developer per active day and $150 to $250 per developer per month across enterprise deployments, with 90% of users staying under $30 per active day. That shape means a small group of heavy users drives most of the bill. Seats suit them and overcharge everyone else, which is exactly why both seat tiers exist and why you should mix them. All of that sizes the bill and none of it sizes the return, which is a separate calculation against the hours the tool gives back at your team's loaded rate, worked through in whether a Claude Code plan pays for itself.

Run a pilot before you commit. Put five people on seats for a month, read the analytics dashboard, then buy for the pattern you measured. The same reasoning applied to one person is covered in what Claude Code costs for a single developer, and the general version of the tradeoff is in API spend versus a subscription for the same work.

If you go the API key route instead, the Claude Console meters everything per token to your organization. When the first member signs in to Claude Code with a Console account, Anthropic creates a workspace called "Claude Code" automatically. Claude Code mints a per-user key there at sign-in, and you cannot create keys in it yourself: it exists so that Claude Code traffic is separable from your production workloads. Set a workspace spend limit there. It is also the only workspace that supports per-user monthly spend limits. Claude Code usage is rate-limited separately, but if your organization has custom rate limits it still counts toward the organization's overall API limits, so set a workspace rate limit too, or a busy afternoon of coding can eat into the headroom your production application needs. Per-user reporting comes from the Console dashboard and the Claude Code Analytics API, using an Admin API key.

What Bedrock and Vertex change

On Amazon Bedrock, Google Cloud and Microsoft Foundry, Claude Code is billed per token to your cloud account and spend controls live in your cloud provider's billing console. The token rates are the ones your cloud contract sets, so this path is chosen for procurement, data residency or an existing committed spend agreement. It is not a discount by default.

The cost of that path is visibility. Claude Code sends no metrics from your cloud back to Anthropic, so the analytics dashboards and the Claude Code Analytics API do not cover this usage at all. To get per-user numbers you export OpenTelemetry metrics from each developer's machine into your own observability stack, or you route traffic through a gateway: Anthropic's self-hosted Claude apps gateway, which also enforces per-user spend limits, or an LLM proxy that tracks spend per key. Both are real work that a seat plan does for you. Decide with the Bedrock and Vertex setup for Claude Code in front of you, because the reporting gap surprises people after the contract is signed.

Where the seat model stops fitting

Mixed sign-in is the quiet one. Each developer is metered according to the method they authenticated with, so half a team on seats and half on an API key produces two bills and two reports that never reconcile. Pick one method per team and check it during onboarding.

Automation is the other. Agent teams spawn several Claude Code instances, each with its own context window, and use roughly seven times the tokens of a standard session when teammates run in plan mode. Scheduled tasks fire on their interval and send the full context each time, even while nobody is at the keyboard. A seat allowance sized for a person is not sized for a person plus four agents, and this is the usual reason a developer hits a weekly limit on a Wednesday. If your usage looks like this, read how Claude Code actually spends tokens before you buy more seats, because context discipline is cheaper than capacity.

FAQ

Is Claude Code included in a Claude Team seat, or does it cost extra?

It is included. Claude Code comes with every Team plan seat, standard and premium, and with the single Enterprise seat on new and self-serve Enterprise plans. Premium seats do not unlock Claude Code, they raise the usage allowance: Anthropic's help centre puts a Team premium seat at 6.25 times the Pro plan's per-session usage allowance. Older Enterprise agreements and HIPAA-ready Enterprise organizations are the exception, because they have separate Chat and Chat plus Claude Code seat types, and a person left on a Chat seat has no access until an owner reassigns them.

Why is the Enterprise seat cheaper than a Team premium seat?

Because it contains no usage. Read on 31 October 2026, the Enterprise seat lists at $20 per user per month billed annually, against $100 for a Team premium seat, and Anthropic's billing article states that every token is billed separately at standard API rates on top of the Enterprise seat fee, or at 1.1x those rates for Opus 4.6, Sonnet 4.6 and later models if your organization turns on US-only inference. The Team premium price is a bundle of access and allowance. The Enterprise price is access only, so the total depends entirely on how much your developers actually run.

How do I see what each developer spends on Claude Code?

On Team and Enterprise, use the spend report in org analytics. It lists estimated spend per user and per model, exports to CSV, refreshes daily and lags by one day. On Team and seat-based Enterprise plans it only shows usage-credit spend once usage credits are enabled. For estimated dollar costs inside the seat allowance, export Claude Code's OpenTelemetry cost metrics with user attribution; these estimates are not additional subscription charges. A current usage-based Enterprise plan exports the organization's full usage. Enterprise plans can pull the same data through the Enterprise Analytics API with a read:analytics key. On Bedrock, Google Cloud or Microsoft Foundry, none of those reports apply, and per-user attribution has to come from OpenTelemetry export or a gateway you run yourself.

When are API keys or a cloud provider cheaper than seats?

When your usage sits well below a seat's price and stays there. A team at the median of $40 a month per person in Claude Code consumption that Anthropic published in August 2026 pays less on metered billing than on $100 premium seats. A team at the mean of $215 pays more than twice as much. Since the documented distribution is skewed, with 90% of users under $30 per active day and a small group far above it, the usual answer is a mix: premium seats for the heavy users, metered access for everyone else. Measure a pilot group for a month before committing, because a guess here compounds across every seat you buy.