How much does Claude Code cost per developer per month in tokens?

Short answer

Anthropic's Claude Code docs give an average of about $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. On a Pro, Max or Team seat you pay a fixed price and hit usage limits instead; with an API key you pay per token.

The numbers Anthropic publishes

The costs page of the Claude Code docs gives three figures for enterprise deployments:

Measure Figure
Average per developer per active day about $13
Average per developer per month $150 to $250
Share of users under $30 per active day 90%

These are averages of API-rate usage. A developer who runs long agent sessions, works on a large codebase or uses Opus for everything will sit above them. Someone who asks a few questions a day will sit far below.

Three ways to pay

  1. API key through the Console. You pay per input and output token. Claude Code creates a workspace called “Claude Code” automatically, so the spend is easy to separate. This is where the $13 per day figure applies directly.
  2. Individual plan. Pro at $20 a month, Max at $100 or Ultra at $200 as of November 2026. You pay a fixed price and run into usage limits instead of a bill: Pro and Max have a 5-hour and a weekly limit, while verified Ultra accounts have no 5-hour limit. Claude chat and Claude Code share those limits.
  3. Team or Enterprise seat. Team Standard seats are $25 a month ($20 billed yearly) and Premium seats $125 ($100 billed yearly), with five times the usage of Standard. Usage comes from a seat allowance with a rolling 5-hour window and a weekly window. Enterprise is $20 per seat billed yearly plus usage at API rates.

When a plan’s limit runs out, you can keep working with usage credits or Console credit at API rates. Watch one thing: if ANTHROPIC_API_KEY is set in your environment, Claude Code bills that key instead of the subscription.

What drives the token bill

The docs point at a few levers:

  1. Model choice. Sonnet for most tasks, Opus for complex ones. Defaulting to the largest model raises cost on every turn.
  2. What loads into every session. CLAUDE.md is read at the start of each session. The docs advise keeping it under 200 lines. Enabled plugins also put their skill names and descriptions in context every turn.
  3. Long sessions. Context grows with every tool result. Clearing between unrelated tasks keeps each turn cheaper.
  4. Re-explaining. Every time a developer explains the same project background again, those tokens are processed again. Prompt caching lowers the API cost of repeated context, but it does not remove it, and on a subscription it eats into the limit.

The misunderstanding: tokens are the whole cost

The token bill is visible. The time spent on context is not. A developer who starts each task by explaining the architecture, the customer constraint or the decision from last week pays twice: in tokens and in minutes. Across a team that repeats, because each developer’s agent learns separately. Auto memory is machine-local, so a lesson one developer taught their agent is not available to the next. Writing stable knowledge into the shared CLAUDE.md or .claude/rules/ helps; see how to share Claude Code context with your team.

Budgeting for a team

For a team of five that uses Claude Code daily, the published averages suggest $750 to $1,250 a month at API rates. Compare that with seats: five Standard seats at $25, or a mix where the two heaviest users get Premium. Then set spend limits per person or group, and look at the spend report after the first month rather than guessing. For the plans themselves, see Claude Code pricing.

Where Bombos fits

Bombos does not lower the token price. It aims at the part of the bill that comes from re-explaining: it is a plugin that gives Claude Code and the team’s other AI tools one shared memory, so context one developer gave once is there for the others, retrieved when it fits the task instead of loaded every turn. One developer can try it on their own agents first.

Related questions

How do I see what I have spent?

Type /usage in Claude Code for your consumption and recommendations, or /status for what is left of your plan's limits. Admins on Team and Enterprise can download the spend report as CSV.

Can I cap spending per developer?

Yes. Admins set spend limits at the organization, group or individual level.

What rate limits should I set for API users?

The docs suggest 200k to 300k tokens per minute per user for teams of 1 to 5, and 100k to 150k for teams of 5 to 20, because not everyone works at the same moment. These are per-user planning figures for organization-level rate limits, not individual spending caps.

Is a subscription cheaper than the API?

A subscription fixes the base fee, but extra usage credits can still add charges, and Enterprise bills usage separately. For someone who works in Claude Code most days a seat is usually the predictable choice; for occasional users, paying per token can be cheaper. Compare against the $13 per active day average.

Sources

  1. Claude Code docs: Manage costs effectively
  2. Claude pricing
  3. Claude Help Center: Using Claude Code with your Pro or Max plan
The first half hour · free

What does your AI forget?

  1. 01You tell us which AI you use and where it keeps starting over.
  2. 02We show what a shared memory changes there, on your own examples.
  3. 03You keep the plan, with or without Bombos.
Piet Baudoin
Piet BaudoinYou talk to Piet, from the Bombos teamDutch team from Utrecht and Nijmegen · data in Europe
We reply the same working day.