GVGanivra

Free calculator · AI coding tools

What AI coding tools cost your team.

Claude Code, Cursor, GitHub Copilot and OpenAI Codex, priced for your team: seats, what each seat includes, what Claude Code costs on per-token billing, and when running an open model on your own GPUs pays off.

Prices checked against each provider's pricing page on . Prices exclude tax.

Calculator

Billing

Seat cost for 25 developers, annual billing

  1. GitHub CopilotCopilot Business or Enterprise: 20 × Business $19, 5 × Enterprise $39
    $575/mo$6,900 a year · $23 per developer
  2. Claude CodeClaude Team: 20 × Standard seat $20, 5 × Premium seat $100
    $900/mo$10,800 a year · $36 per developer
  3. OpenAI CodexChatGPT Business: 20 × Standard seat $20, 5 × Premium seat $100
    $900/mo$10,800 a year · $36 per developer
  4. CursorCursor Teams: 20 × Premium seat $30, 5 × Super seat $100
    $1,100/mo$13,200 a year · $44 per developer

Seat prices only. Heavy users can pass the included usage on every tool, and anything past it is billed on top.

Or pay per token

Claude Code on API billing

No seats. Anthropic reports an average of about $13 per developer per active day across enterprise deployments, and under $30 a day for 90% of users.

$4,875/mo$58,500 a year · $195 per developer

Or run your own

Self-host an open model

Rent GPUs, run an open coding model such as GLM, and keep a hosted model for the hardest work. GPUs bill around the clock whether anyone is coding or not, and B200s rent in 8-GPU nodes, so small teams pay for capacity they don't use.

Already running GPUs? See the cost per call and customer on them
$55,045/mo8 GPUs $39,070 + people $15,000 + API still used $975

$50,170 a month more than all-API ($4,875)

First costs less than the API at about 848 developers

Team plans

Seat prices, and what each seat includes.

Claude Code

Claude Team
Standard seat
$20 annual · $25 monthly
Premium seat
$100 annual · $125 monthly

Included. Claude Code and Claude chat share each seat's allowance, which resets every 5 hours and every week. A Premium seat has 5x the usage of a Standard seat.

Past the allowance. Admins can turn on usage credits so developers keep working past the allowance, with spend limits per organization, group or member.

Pricing source →

Cursor

Cursor Teams
Premium seat
$30
Super seat
$100

Included. Each seat has its own weekly pool for Cursor models. Super has 3.75x the Premium pool; an Ultra seat at $200 has 12.5x.

Past the allowance. After the pool, usage draws from a prepaid team balance at token rates. Third-party models are billed at API list price plus $0.25 per million tokens.

Pricing source →

GitHub Copilot

Copilot Business or Enterprise
Business
$19
Enterprise
$39

Included. Business includes 1,900 GitHub AI Credits per user a month and Enterprise 3,900, pooled across the organization. One credit is $0.01.

Past the allowance. Usage beyond the pool costs $0.01 per credit. Budgets can be set per organization, enterprise or cost center.

Pricing source →

OpenAI Codex

ChatGPT Business
Standard seat
$20 annual · $25 monthly
Premium seat
$100 annual · $125 monthly

Included. Standard seats include ChatGPT and Codex within plan limits. Minimum of 2 seats.

Past the allowance. After the included limits, eligible usage continues from credits the workspace buys.

Pricing source →

Per user per month. Where only one price is shown, the provider doesn't publish an annual rate. Enterprise plans for each tool are priced by sales.

Read the bill carefully

The seat price is the floor, not the cost.

  1. 01

    Allowances run out

    Every seat comes with an allowance: Claude resets every 5 hours and weekly, Cursor weekly, Copilot monthly. Past it, usage credits or a prepaid balance take over, billed on top of the seat.

  2. 02

    Heavy users cost more than the average

    Anthropic reports about $13 per developer per active day on API billing, and under $30 a day for 90% of users. A developer running agents all day can use more in a week than a Standard seat costs in a month.

  3. 03

    Invoices show seats, not people

    The bill says how many seats you pay for. It doesn't say who used them, which seats sit idle, or which team ran up the overage.

Sources: Claude Code averages from Anthropic, Manage costs effectively; B200 price from Lambda on-demand pricing.

Per developer

How to see what each developer really costs.

Each tool reports on itself. None of them shows one figure across tools, by team, at what you actually pay. For Claude Code, follow the guide to tracking cost per developer.

  • Claude Team and EnterpriseA spend report in org analytics, with CSV export per user and model. It covers usage-credit spend only: usage inside the seat allowance isn't shown in dollars. Enterprise adds an analytics API.
  • Claude on API billingThe Console cost page and the Claude Code Analytics API, with daily metrics per user.
  • Cursor TeamsThe team dashboard and the Admin API, with daily usage and spend per member.
  • GitHub CopilotA downloadable usage report per user. Per-user analytics are limited to enterprise owners and billing managers.

Questions

AI coding tool costs FAQ

How much does Claude Code cost per developer?

On the Claude Team plan, a Standard seat is $20 a month billed annually or $25 monthly, and a Premium seat with 5x the usage is $100 or $125. On API billing there is no seat: Anthropic reports an average of about $13 per developer per active day and $150 to $250 a month across enterprise deployments.

Which AI coding tool is cheapest for a team?

By seat price, GitHub Copilot Business at $19 per user a month. Seat prices aren't the whole cost, though: each tool includes a different amount of usage, and heavy users pay for usage past the allowance on every tool. Compare what is included, not only the seat.

Should we buy seats or pay per token?

Seats give a predictable bill with a usage ceiling. Per-token billing has no ceiling but costs only what is used, which suits teams with a few heavy users and many light ones. Many teams start on seats and give usage credits to the developers who hit the limits.

Is self-hosting an open coding model cheaper?

At large scale it can be. Revolut reported moving much of its coding work to GLM on 56 B200 GPUs for about 2,000 engineers. GPUs bill around the clock and rent in 8-GPU nodes, so a small team pays for idle capacity, plus the people who keep the model running. Put your own team size into the self-hosting box to see where it crosses the API cost.

Does the calculator include overage?

No. It shows seat prices, plus a separate per-token estimate for Claude Code. Overage depends on how each developer works, so measure a few weeks of real usage before you budget for it.

How do we see what each developer actually costs?

Each tool has its own report, covering only that tool. Claude Code can also export OpenTelemetry from every developer's machine, which gives per-request token and cost data by person and team in near real time, whatever plan you are on. Ganivra accepts that export directly.

See Claude Code spend per developer, from real usage.

Point Claude Code's OpenTelemetry export at Ganivra and get cost by developer, team and model. 14-day free trial, no credit card, from $29 a month after.

Start free trialRead the setup guide