Claude Code, Codex, Cursor, and Copilot: What AI Coding Tools Cost per Developer in 2026
Team pricing for the four main AI coding tools, what a 50-person team pays, why the bill spikes, where each shows cost per developer, and how to spend fewer tokens.
Team seats for AI coding tools cost $19 to $200 a developer a month, but usage decides the bill. Anthropic puts Claude Code at about $13 per developer per active day, or $150 to $250 a month, and every tool here bills usage beyond the seat at API rates or in credits. To keep the bill in line, give each developer the plan their usage needs, set spend limits per person, read each tool's per-developer report every week, and cut the tokens each session sends.
Below: what Claude Code, OpenAI Codex, Cursor, and GitHub Copilot charge teams, what a 50-person team pays, why the bill spikes, where each tool shows cost per developer, and each vendor's own advice for spending fewer tokens.
What does each AI coding tool cost per seat?
| Tool | Team plans | What the seat includes | Usage beyond the seat |
|---|---|---|---|
| Claude Code | Team Standard $20 a seat a month billed annually ($25 monthly); Team Premium $100 ($125 monthly); Enterprise $20 a seat billed annually, 20 seats minimum | Claude, Claude Code, and a usage allowance; Premium has 5 times Standard's | Usage credits you turn on, at standard API rates; Enterprise bills all usage at API rates |
| OpenAI Codex | ChatGPT Business $20 a user billed annually ($25 monthly); Enterprise by quote; or an API key at API prices | Codex in ChatGPT, with message limits per 5 hours by model | Credits at a price set by your plan or agreement |
| Cursor | Teams seats at $30, $100, and $200 a month; Enterprise by quote | A weekly pool for Cursor's own models | Claude, GPT, and Gemini at API list price from prepaid credit, plus a $0.25 per million token rate on eligible requests |
| GitHub Copilot | Business $19 a seat a month; Enterprise $39 | 1,900 AI Credits (Business) or 3,900 (Enterprise) at $0.01 each, pooled across the organization; code completions unlimited | $0.01 a credit, on by default |
From each vendor's own pricing pages and docs, checked October 24, 2026. [Claude Team pricing](https://claude.com/pricing/team) · [Codex pricing](https://learn.chatgpt.com/docs/pricing) · [Cursor pricing](https://cursor.com/pricing) · [Copilot billing](https://docs.github.com/en/copilot/concepts/billing-and-usage/organizations-and-enterprises/billing).
Claude Code
Claude Code comes with every paid Claude plan. Individuals pay $20 a month for Pro ($17 billed annually) or from $100 for Max, with 5 or 20 times Pro's usage. All plans limit use in a rolling five-hour window, and paid plans add weekly limits; past them, you can turn on usage credits at standard API rates. Claude pricing. Enterprise charges a $20 seat plus all usage at API rates, where Claude Sonnet 5.5 costs $2 per million input tokens and $10 per million output tokens.
OpenAI Codex
Codex comes with ChatGPT plans: Plus at $20, Pro from $100, and Business at $20 a user billed annually. On Plus and standard Business seats, OpenAI allows a range of local messages every five hours by model, from 5 to 45 on GPT-6 Astra to 350 to 3,000 on GPT-6 Luna, and weekly limits may apply. Fast mode uses 2.5 times the allowance. You can also run Codex on an API key at API prices, without its cloud features. Codex pricing.
Cursor
Cursor's individual plans run from Plus at $8, limited to its own and Grok models, through Premium at $30, Super at $100, and Ultra at $200, with 4, 15, and 50 times Plus's usage. Each plan has a weekly pool for Cursor's models. Claude, GPT, and Gemini sit outside the pool and bill at API list price from prepaid credit, and Teams also pay a Cursor Token Rate of $0.25 per million tokens on eligible third-party requests. Cursor models and pricing · Teams pricing.
GitHub Copilot
Since June 1, 2026, Copilot bills in AI Credits worth $0.01 each, replacing premium requests. A Business seat costs $19 and includes 1,900 credits, an Enterprise seat $39 with 3,900, and the credits pool across the billing entity, so light users cover heavy ones. Extra credits cost $0.01 each, overage is on by default, and requests keep the model you picked, billed as overage. Code completions stay unlimited. GitHub's announcement · Copilot billing.
What does a 50-person team pay?
Anthropic's own figure for Claude Code is "around $13 per developer per active day and $150-250 per developer per month", and 90% of users stay under $30 per active day. Claude Code costs. Two worked examples for 50 developers:
| Plan | Seats | Usage | Month |
|---|---|---|---|
| Claude Enterprise, 50 seats, usage at API rates | $1,000 | $7,500 to $12,500 at Anthropic's $150 to $250 per developer | $8,500 to $13,500 |
| Copilot Business, 50 seats | $950, with 95,000 pooled credits | 140,000 credits used, 45,000 over the pool at $0.01 | $1,400 |
From each vendor's own pricing pages and docs, checked October 24, 2026. The Copilot usage figure is an example, not GitHub data.
The seat is the small part of the bill. In both cases what moves the month is how many tokens each developer's sessions send, which model they use, and whether anyone sees the spend before the invoice.
Why does the AI coding bill spike?
- Long sessions. Every request resends the conversation, so a session that never clears gets more expensive with each step.
- Large instruction files and many MCP servers.
CLAUDE.md,AGENTS.md, and tool definitions load into every request; Anthropic suggests keepingCLAUDE.mdunder 200 lines and disabling unused servers. - The biggest model by default. GPT-6 Astra allows 5 to 45 messages per five hours on a Plus seat where GPT-6 Luna allows 350 to 3,000, and Claude Opus 5.5 costs twice Sonnet 5.5 per token.
- Fast modes and long contexts: Codex's Fast mode uses 2.5 times the allowance, Cursor bills Fast mode at 2 times, and Grok 4.7 above 256,000 tokens of context bills at 2 times in Cursor.
- Parallel agents. Anthropic says Claude Code's experimental agent teams use about 7 times the tokens in plan mode.
- Overage left on with no budget set. Copilot turns overage on by default.
Where does each tool show cost per developer?
| Tool | Where | Per developer | Lag and range | Scope |
|---|---|---|---|---|
| Claude Code, Team and Enterprise | Spend report (CSV) | Net and gross dollars per person, per model | Daily, 1 day behind, 90 days | All Claude use; usage-credit spend on seat plans, full usage on usage-based Enterprise |
| Claude Code on the Claude Console | Claude Code Analytics API | Tokens, estimated cost by model, sessions, lines, commits, pull requests | About 1 hour behind, one day per call | API usage on the Console; subscription usage and usage through Bedrock or Google Cloud report elsewhere |
| OpenAI Codex | Admin API Codex usage endpoint; Usage Insights by use case | Codex usage per day, per user or for the workspace | Daily; delay not stated | The endpoint moves to the Unified Daily Usage API on December 15, 2026 |
| Cursor | Admin API, Usage Analytics | Spending and usage events per member | Usage Analytics filters up to 10 users and 90 days | Analytics API on Enterprise |
| GitHub Copilot | AI usage report, impact dashboard | Tokens per user, per model; estimated average monthly cost per developer by adoption group | Up to 31 days per report | Detailed report from the billing pages, not the REST API |
From each vendor's own pricing pages and docs, checked October 24, 2026. [Claude spend report](https://support.claude.com/en/articles/12883420) · [Claude Code Analytics API](https://platform.claude.com/docs/en/manage-claude/claude-code-analytics-api) · [Codex Admin API](https://chatgpt.com/public/admin/api-reference) · [Codex analytics](https://learn.chatgpt.com/docs/enterprise/workspace-analytics) · [Cursor analytics](https://cursor.com/docs/account/teams/analytics) · [Copilot billing reports](https://docs.github.com/en/billing/reference/billing-reports) · [Copilot impact dashboard](https://github.blog/changelog/2026-08-07-copilot-impact-dashboard-adds-a-return-on-investment-section).
Claude Code running through Bedrock or Google Cloud doesn't appear in Anthropic's analytics API; Anthropic points those teams to OpenTelemetry or a gateway instead. No single report covers all four tools, so a team that uses two of them reads two reports with different lags and different definitions of a developer's cost.
How do you cap spend per developer?
- GitHub Copilot: budgets per user (one limit for everyone, per cost center, or per person with an expiry date), per cost center, per organization, and per enterprise. A user budget stops usage when it runs out. Copilot budgets.
- OpenAI Codex: user limits as a workspace default, per group, or per person, plus an overage limit. OpenAI notes that alerts "don't stop spending"; limits do. Codex usage and cost.
- Cursor: the Admin API sets a spend limit per user, and Teams have a monthly limit on additional usage.
- Claude Code on the Console: the Claude Code workspace is the only workspace that supports per-user monthly spend limits. Claude workspaces.
How do developers spend fewer tokens?
Each vendor publishes its own advice. The common thread: start fresh for each task, keep standing context small, and use the cheapest model that does the job.
- Claude Code:
/clearbetween tasks and/compactwith instructions; Sonnet over Opus, and Haiku for subagents;/contextto see what fills the window; CLIs over MCP servers where both work; workflow text moved fromCLAUDE.mdinto skills; lower effort for routine work; plan mode and specific prompts./usageestimates spend at list price. Claude Code costs. - OpenAI Codex: trim context, limit sources, set an output length, keep
AGENTS.mdsmall or nested, use fewer MCP servers, and check the usage dashboard every week or two. Codex pricing FAQ. - Cursor: the model you pick sets the burn rate; Fast mode and Grok 4.7 contexts over 256,000 tokens bill at twice the rate. Cursor models and pricing.
- GitHub Copilot: pick the model per task or let auto selection choose, send subagents to cheaper models, start a new conversation per problem, keep the model and reasoning level fixed so the cache holds, and set session credit limits. Optimize Copilot usage.
Common mistakes with AI coding tool costs
- Buying the top plan for everyone, when most developers' usage fits the standard seat.
- Leaving overage on with no budget, so the first sign of a heavy month is the invoice.
- Reading only the seat count. The seat is a fraction of the bill once usage runs at API rates.
- Letting each tool's report live with a different admin, so nobody sees one developer's total across tools.
- Treating the vendor's average as your forecast instead of measuring your own team for a month.
LevelFour reads your Anthropic, OpenAI, and Cursor bills, with Claude Code and Codex inside them, next to the rest of your cloud and SaaS spend, and allocates them by person and team with virtual tags. See what LevelFour can save you in 5 minutes, with read-only access.
Frequently asked questions
How much does Claude Code cost per developer?
Anthropic puts it at around $13 per developer per active day, or $150 to $250 per developer per month, with 90% of users under $30 per active day. On Claude Team the seat is $20 a month billed annually, or $100 for Premium with five times the usage; on Enterprise the seat is $20 plus usage at API rates. Claude Code costs.
Is Claude Code included in a Claude subscription?
Yes. Every paid Claude plan includes Claude Code, within a rolling five-hour usage window and, on paid plans, weekly limits. Past them, you can turn on usage credits at standard API rates.
What are GitHub Copilot AI Credits?
Since June 1, 2026, Copilot bills usage in AI Credits worth $0.01 each, replacing premium requests. Business seats include 1,900 credits and Enterprise seats 3,900, pooled across the organization, with extra credits at $0.01 and overage on by default. Annual Pro and Pro+ subscribers on legacy billing keep premium requests.
How does Cursor charge for Claude and GPT models?
Outside its weekly pool for Cursor's own models, Cursor bills Claude, GPT, and Gemini at API list price from prepaid credit. Teams also pay a Cursor Token Rate of $0.25 per million tokens on eligible third-party requests.
How do I see Claude Code cost per developer?
On Team and Enterprise, download the spend report: net and gross dollars per person and model, daily, one day behind. On the Claude Console, the Claude Code Analytics API returns tokens and estimated cost per user about an hour behind for API usage; subscription usage and usage through Bedrock or Google Cloud report elsewhere.
Which AI coding tool is cheapest for a team?
By seat price, Copilot Business at $19 with 1,900 credits included. By total cost it depends on usage: every tool here bills heavy use at API rates or in credits, so measure a month of your own team's usage before you compare.