Claude Code vs Codex vs Cursor cost per engineer, for teams.
Sources last reviewed
Billed annually, a team seat costs $20 a month for Claude Code on Claude Team, $20 for Codex on ChatGPT Business and $32 for Cursor on Teams. Past a seat's included usage all three can bill by the token, so the model and the work set what each engineer costs.[1][2][3]
Of the 866 agent-written pull requests we merged from August 12 to October 4, 2026, Claude Code opened 822, Codex 31 and Cursor none. One engineer's Claude Code month cost more than a hundred times Anthropic's per-developer average at list prices.
Measure all of them with Zest:
Claude Code, Codex and Cursor sessions side by side, per engineer
The model, skills and MCP tools behind each session
Claude CodeFable 5.1 (max) (with fallback)70.4 · Coding Agent Index
CodexGPT-6 Astra (max)67.0 · Coding Agent Index
Muse CodeMuse Spark 1.3 (xhigh)64.2 · Coding Agent Index
Grok BuildGrok 4.5 (high)64.1 · Coding Agent Index
Kimi Code CLIKimi K362.6 · Coding Agent Index
OpencodeGemini 3.8 Flash (high)61.2 · Coding Agent Index
Cursor CLIComposer 2.5 Fast38.3 · Coding Agent Index
Demo setups use current model families. The pairings above are the combinations benchmarked by Artificial Analysis; rankings measure performance, not popularity.
Illustrative daily trends · Other totals: last 30 days. Build Hours: this week, model-estimated activity including overlap; not engineer hours.
USED BY ENGINEERING TEAMS AT
AT A GLANCE
Claude Code vs Codex vs Cursor, side by side
What each tool documents, row by row.
Claude Code vs Codex vs Cursor, side by side
Claude Code
Codex
Cursor
Team seats for 5 engineers
$100/mo — Claude Team Standard at $20 a seat, billed annually ($25 monthly).[1]
$100/mo — ChatGPT Business at $20 a seat, billed annually ($25 monthly).[2]
$160/mo — Cursor Teams Standard at $32 a seat, billed annually ($40 monthly).[3]
Where it runs
Terminal CLI, VS Code and JetBrains extensions, a desktop app, the web and mobile, plus Slack and GitHub Actions.[4]
Codex CLI, an IDE extension (VS Code, Cursor, Windsurf), the ChatGPT desktop app, Codex Cloud and @codex review on GitHub.[5]
The Cursor editor (Agent pane and Agents Window), the Cursor CLI, Cloud Agents in VMs, and JetBrains through ACP.[6]
Models
Claude models only (Fable, Opus, Sonnet, Haiku); Anthropic doesn't support routing Claude Code to other models.[7]
OpenAI GPT models; local clients can point at a custom Responses-API provider such as Ollama.[8][9]
OpenAI, Anthropic, Google, SpaceXAI (Grok) and more, plus Cursor's own Composer; Auto routing and bring-your-own keys.[10]
Team analytics, built in
Team and Enterprise dashboard: daily users, sessions, lines accepted, accept rate and PRs with Claude Code; API on Enterprise.[11]
Workspace analytics of messages, credits and active users; a Codex dashboard and Analytics API on Enterprise.[12][13]
Teams dashboard: AI share of committed code, agent edits, messages by model, daily users and a leaderboard.[14]
Checking your own usage
/usage (also /cost and /stats) shows plan usage, session cost and a 24h/7d breakdown by skill, subagent and MCP server.[15]
A usage dashboard with limits and reset times; /status and /usage daily, weekly or cumulative in the CLI.[2][16]
A usage dashboard and editor settings show included usage per pool, Cursor models and other models.[10]
Each tool's column is read from its vendor's own docs and pricing page, dated in the sources below.
Claude Code, Codex or Cursor: which one fits your team
Start from the models you want and the plan you already pay for: Claude models on a Claude plan, GPT models on a ChatGPT plan, or Cursor's editor with several labs' models on its own plan.
Claude Code's and Codex's extensions install in Cursor, and Cursor reads CLAUDE.md and AGENTS.md and loads both tools' skills. Each extension still bills to its own vendor's plan.
One change to plan for: on August 28, 2026, OpenAI said it intends to stop providing its models to Cursor, with a proposed shutoff date of November 12. Cursor still lists them today.
What a team seat costs on each, the bigger seat for heavy users, and what each bills once a seat's included usage runs out:
Team plans, per seat a month before tax. Prices verified October 12, 2026 against each vendor's pricing page, OpenAI's Help Center and Anthropic's and Cursor's docs, listed in the sources below.
Claude Code
Codex
Cursor
Seat price
$25 a seat a month, or $20 billed annually
$25 per user a month, or $20 billed annually
$40 per user a month ($32 billed yearly)
Seat for heavy users
Team Premium, $125 a seat a month, or $100 billed annually: 5x a Standard seat's usage
Business Premium, $125 per user a month, or $100 billed annually: 5x the usage, no five-hour limit
Teams Premium, $120 per user a month ($96 yearly): 5x a Standard seat's usage
Included usage
Per member, five-hour and weekly limits; no published count
Per seat, an estimated number of messages per five hours, plus weekly limits
Per seat, in two pools (Cursor's models, others); no published count
Past the included usage
Usage credits at API rates, if an owner turns them on
Workspace credits, one pool for all members, or an API key at API rates
On-demand usage at API rates, on by default
On top of the model's price
Nothing
Nothing
$0.25 per million tokens on third-party models, included usage too
Spend limits
For the organization and each member
Monthly credit limits per seat type and per user; none by default
One monthly limit for the team; per member on Enterprise
What the vendor says a developer spends
$150–250 per developer a month, about $13 per active day (Anthropic's average)
No dollar figure; OpenAI estimates messages per five hours
$60–$100 a month for daily Agent users, $200 or more for power users (Cursor)
Why the seat price isn't the cost per engineer
A seat is the floor. Its included usage belongs to one person on all three tools, so the engineers who run agents all day hit the limit first, then wait for the reset or pay by the token.
Past the seat the meter is tokens, priced per model: at list prices OpenAI's GPT-6.1 Sol costs half what Anthropic's Opus 5.5 does per input, cached and output token. Cursor charges the labs' API prices plus its token rate on Teams and Enterprise; its own Composer and Grok pay no token rate.
A price per token isn't a price per task: agents spend different numbers of tokens on the same work. Compare them on your own merged pull requests.
What one engineer's Claude Code month cost at list price
One engineer's 318 Claude Code sessions (September 4 to October 3, 2026) used 67.9 billion tokens, 98.4% of them cache reads, and led to 601 merged pull requests. At Anthropic's API list prices on October 4, 2026, the month came to more than a hundred times the top of Anthropic's per-developer average. That is list price, not the bill: the engineer pays for Max 20x and usage credits (Claude Code pricing has the table).
Cache reads were 57% of that cost and subagents 34%. On Cursor Teams the token rate applies to cached tokens too, more than doubling what a cache read of Opus 5.5 costs.
How to see each engineer's cost across all three
Each vendor's admin view covers only its own agent, mostly the spend past the seat: Claude Team shows each member's usage-credit spend, ChatGPT Business each member's credits and messages, and Cursor each member's on-demand spend. None shows the other two tools, or which sessions led to merged pull requests.
Zest prices each Claude Code and Codex session at list rates, inside the allowance too, per engineer; for Cursor it reads tokens and cost from Cursor's usage dashboard once the engineer turns that on. Each session is joined to the pull request it led to, so cost per merged pull request compares across the three (Claude Code usage, Codex usage, Cursor usage). Zest reads no invoices: its numbers are list prices, not your bill.
TRY IT ON YOUR TEAM
See what your team ships with AI.
Starter is free (3 seats, no credit card). Pro is $49 a month with 5 seats and Max is $149 a month with 10 seats — priced per workspace, not per engineer, with extra seats at $19 a month.
Which is cheapest for a team: Claude Code, Codex or Cursor?+
On Standard seats, Claude Team and ChatGPT Business list the same price and Cursor Teams more; on the heavy-user Premium seats, Cursor lists lowest. Past the seat it depends on the model and the work, so compare cost per merged pull request on your own team before you renew.
Are any of them free?+
Codex is, in a limited way: ChatGPT Free and Go run GPT-6 Luna in the desktop app, subject to rollout. Cursor's Hobby plan is free with limited Agent requests. Claude Code needs a paid Claude plan or an API key.
Can one subscription cover all three?+
No. Each vendor bills its own plan. Cursor runs other labs' models on a Cursor plan, and Claude Code's and Codex's extensions run inside Cursor, but they bill to their own vendor's plan or API key.
Do all three charge API rates past the limit?+
Close to it. Claude's usage credits bill at standard API rates. Cursor bills on-demand usage at API rates and adds its token rate on Teams and Enterprise for third-party models. Codex charges workspace credits per million tokens at a rate for each model, and OpenAI says what a credit costs depends on the plan or agreement.
Winding / Teams
