Skip to content
AI Developer Toolkit

Find your level — free Buy the Toolkit Buy Toolkit Buy the Toolkit Buy Toolkit Open the guides Read
Navigation menu
Buy the Toolkit Find your level — free Buy the Toolkit Open the guides Sign in

AI coding tool plan prices and how usage is billed

PL

Level 1 Reference Updated

Claude Code is included in Claude Pro ($20 a month), Max ($100 or $200) and Team seats (from $20, billed annually). Codex is included in ChatGPT Plus ($20), Pro ($100, $200 or $500), Business, Edu and Enterprise. Cursor individual plans cost $20 to $200 a month and GitHub Copilot $10 to $100. Every plan meters agent work through usage windows or credits, so the seat price is a floor.

Finance wants one number for eight engineers before the renewal. One engineer already pays for Max 20x on a personal card, and the comparison someone forwarded quotes prices nobody can source. This page is for the tech lead or CTO who signs the seat order and for the developer who is choosing a personal plan. It keeps plan prices only. Return on investment is measured as cost per accepted change, and per-token model prices live on the models hub.

What you get from this pricing comparison

Section titled “What you get from this pricing comparison”
  • A dated table of which plans include Claude Code, Codex, GitHub Copilot, and Cursor, with the price and the source for every row.
  • How each tool meters agent work after the seat price: usage windows, usage credits, AI credits, and API keys.
  • A six-step worksheet that prices a team from its own invoices, with a CSV template and a worked example.
  • Three copy-paste prompts: build the worksheet from billing exports, size one developer’s plan from measured usage, and normalise enterprise quotes.
  • The checks that prove the estimate without auditing every invoice line, and what breaks in a price comparison.

Which plans include Claude Code, Codex, and Copilot?

Section titled “Which plans include Claude Code, Codex, and Copilot?”
ProductPlanList price (USD)Agent access and what the plan addsSource
ClaudeFree$0No Claude Codeclaude.com/pricing
ClaudePro$20/month, or $17/month billed annually ($200 up front)Claude Code includedclaude.com/pricing
ClaudeMax 5x$100/monthClaude Code; five times Pro usage; $100 of monthly Claude API credits since 2026-10-07 (not for interactive Claude Code; Anthropic, 2026-10-12)Max plan, API credits
ClaudeMax 20x$200/monthClaude Code; twenty times Pro usage; $200 of monthly Claude API credits since 2026-10-07 (not for interactive Claude Code; Anthropic, 2026-10-12)Max plan, API credits
ClaudeTeam Standard$25/seat monthly, $20/seat annualClaude Code and Claude Cowork; $20 of pooled monthly Claude API credits per seat since 2026-10-07 (not for interactive Claude Code; Anthropic, 2026-10-12)claude.com/pricing, API credits
ClaudeTeam Premium$125/seat monthly, $100/seat annual“5x more usage than standard seats”; $100 of pooled monthly Claude API credits per seat since 2026-10-07 (not for interactive Claude Code; Anthropic, 2026-10-12)claude.com/pricing, API credits
ClaudeEnterprise$20/seat/month billed annually, plus usage at API ratesClaude Code; usage cost “scales with model and task”claude.com/pricing
ChatGPTFree$0 (OpenAI, 2026-10-12)GPT-6 Luna at Standard speed in the desktop app, subject to rolloutOpenAI Codex pricing
ChatGPTGo$8/month (OpenAI, 2026-10-12)GPT-6 Luna at Standard speed in the desktop app, subject to rolloutOpenAI Codex pricing
ChatGPTPlus$20/month (OpenAI, 2026-10-12)Codex on the web, CLI, IDE and iOS; cloud integrations; GPT-6.1 Sol and GPT-6 Luna; creditsOpenAI Codex pricing
ChatGPTPro$100/month (OpenAI, 2026-10-12)Codex; currently no five-hour limit; 5x Plus usage on Codex (secondary: TechCrunch, 2026-04-09)OpenAI Codex pricing, TechCrunch
ChatGPTPro$200/month (OpenAI, 2026-10-12)Codex; currently no five-hour limit; 20x Plus usage on Codex (secondary: TechCrunch, 2026-04-09)OpenAI Codex pricing, TechCrunch
ChatGPTPro$500/month (OpenAI, 2026-10-12)Codex; currently no five-hour limit; the Ultrafast speed tier is available only on Pro $500 and eligible Enterprise or Edu plansOpenAI Codex pricing
ChatGPTBusiness$20/user/month billed annually, $25/user/month billed monthly; 2+ users (OpenAI, 2026-10-12)Codex includedOpenAI Codex pricing
ChatGPTEdu, EnterpriseContact sales (OpenAI, 2026-10-12)Codex included; with flexible pricing there are no fixed rate limits, without it the same per-seat usage limits as Plus for most featuresOpenAI Codex pricing
GitHub CopilotFree$0Auto model only; 2,000 completions/monthCopilot plans
GitHub CopilotPro$10/month (GitHub docs, 2026-10-12)1,000 base AI credits, plus 500 flexCopilot billing
GitHub CopilotPro+$39/month (GitHub docs, 2026-10-12)3,900 base AI credits, plus 3,100 flexCopilot billing, as above
GitHub CopilotMax$100/month (GitHub docs, 2026-10-12)10,000 base AI credits, plus 10,000 flexCopilot billing, as above
GitHub CopilotBusiness$19/seat/month (GitHub docs, 2026-10-12)1,900 base AI credits per usercopilot.yml
GitHub CopilotEnterprise$39/seat/month (GitHub docs, 2026-10-12)3,900 base AI credits per usercopilot.yml
CursorHobby$0 (Cursor, 2026-10-12)Limited Agent requests; Composer; Agent, Chat and Tab with the Auto modelcursor.com/pricing, Cursor help
CursorPro$20/month (Cursor, 2026-10-12)Unlimited Tab completions, extended Agent usage limits on all models, Bugbot and Cloud Agentscursor.com/pricing, Cursor help
CursorPro+$60/month (Cursor, 2026-10-12)The same individual-plan features as Pro; Cursor recommends Pro+ for daily Agent userscursor.com/pricing, Cursor help
CursorUltra$200/month (Cursor, 2026-10-12)The same individual-plan features as Pro; Cursor recommends Ultra for Agent power userscursor.com/pricing, Cursor help
CursorStart (India only)₹649/month, tax inclusive (not USD) (Cursor, 2026-10-12)The Cursor Models pool only, with Cloud Agents and Cursor for iOS; no Other Models pool, on-demand usage, Bugbot, Auto, Automations or Cursor SDKcursor.com/pricing, Cursor help
CursorTeams Standard$40/user/month, $32 billed annually (Cursor, 2026-10-12)Usage counted per user and not transferable; on-demand usage on by defaultcursor.com/pricing, Cursor help
CursorTeams Premium$120/user/month, $96 billed annually (Cursor, 2026-10-12)“5x the usage of a Standard seat”cursor.com/pricing, Cursor help
CursorEnterpriseCustom (Cursor, 2026-10-12)Contact Cursorcursor.com/pricing

The “5x” and “20x” labels are ratios, not message counts. Anthropic describes plan usage as rolling five-hour and weekly windows, not as a number of Claude Code messages. Treat OpenAI’s Pro ratios the same way: how far a plan goes depends on the model, the effort level, the context size, and how many agents you run at once.

Anthropic’s monthly API credits for Max and Team started on 2026-10-07. Team credits are pooled across seats and capped at $500, credits do not roll over, and Free, Pro and Enterprise plans are not eligible. They pay for Claude Platform work (the Claude API, Batches, Managed Agents and the Agent SDK), not for interactive Claude Code in the terminal, IDE, desktop or web, and they do not change your usage limits (Anthropic support, read 2026-10-12). ChatGPT Work and Codex share usage, and Pro plans currently have no five-hour limit (OpenAI, 2026-10-12).

No included-usage dollar amount per Cursor plan is verified here. Figures from search snippets or older articles, including earlier versions of this page, are not verified. For what each tool does rather than what it costs, see the capability comparison.

How is agent usage metered beyond the seat price?

Section titled “How is agent usage metered beyond the seat price?”

The seat buys a usage allowance. What you pay on top depends on the tool, and per-token rates for API keys and usage credits are on the models hub.

Allowance. Cursor’s included-usage dollar amounts per plan are not verified here. Cursor’s own guidance is that daily Agent users typically use $60 to $100 a month in total and power users often $200 or more (cursor.com, read 2026-10-12). Cursor says usage beyond the plan continues at the same API rates with pay-as-you-go billing, and that all Auto modes bill at the list price of the model each request is routed to. On Teams and Enterprise, third-party requests, including Auto routes and bring-your-own-key use, also incur the Cursor Token Rate of $0.25 per million tokens (cursor.com, read 2026-10-12).

What bills on top.

  • Bugbot bills by usage on Teams and individual plans, replacing a $40-per-seat subscription. Cursor puts the average run at $1.00 to $1.50, depending on pull request size and complexity; Teams bill it from on-demand usage, individuals from included usage and then on-demand (Cursor help, read 2026-10-12).
  • Max Mode exists only on legacy request-based plans, billed at the model’s API rate plus 20% (Cursor help, read 2026-10-12).

Check it. Read the usage page in your Cursor dashboard for one full billing cycle before you change plans, and write down the date you read it. To pull Cursor usage onto your own machine, the agent cost tracking guide shows how tokscale reads Cursor’s usage export.

How do you price a team from its own invoices?

Section titled “How do you price a team from its own invoices?”

List prices tell you the floor. The number finance needs comes from one billing cycle of real invoices. Here is the order that works:

  1. Count named seats. One seat per person on a team or enterprise plan. Consumer logins must never be shared; a shared account is not a cost saving, and it removes the per-person audit trail.

  2. Record each price with its source and date. One row per seat type and plan, in the CSV template below. A row without a source URL and a checked_on date is not a price, it is a guess.

  3. Pull one full billing cycle of variable spend. Usage credits, extra AI credits, and on-demand charges come from each vendor’s billing page. For per-developer detail, npx ccusage@latest monthly --json totals Claude Code, Codex, and other agents’ local logs at list price. Use it to split cost between people, never as the bill.

  4. Add the work billed outside the seats. CI jobs on API keys (claude -p, codex exec), managed review runs, fast mode, and any gateway spend. These arrive on a different invoice and are the line most often missed.

  5. Add it up. Keep fixed and variable spend in separate columns so a renewal discussion can move one without hiding the other:

    monthly total = Σ (seats × seat price)             # fixed
                  + usage credits and extra AI credits # variable, from invoices
                  + API-key and CI spend               # variable, from invoices
                  + tax, support, and platform fees

  6. Set a re-check date. Re-open every price source before a purchase decision, and at least every 14 days while a decision is open. Vendors changed models, limits, and billing several times in 2026.

The worksheet template, one row per invoice line or seat type. The first rows show a team of eight on Claude Team with two Premium seats, plus Copilot Business for the same people:

item,vendor,plan,unit_price_usd,units,billing,source,checked_on,kind
claude-team-standard,Anthropic,Team Standard,25,6,monthly,https://claude.com/pricing,2026-09-26,fixed
claude-team-premium,Anthropic,Team Premium,125,2,monthly,https://claude.com/pricing,2026-09-26,fixed
copilot-business,GitHub,Business,19,8,monthly,https://github.com/github/docs/blob/main/content/copilot/get-started/plans.md,2026-09-26,fixed
claude-usage-credits,Anthropic,usage credits,MEASURED,1,monthly,invoice-2026-09.pdf,2026-09-26,variable
ci-api-key,Anthropic,API,MEASURED,1,monthly,console-export-2026-09.csv,2026-09-26,variable

Replace each MEASURED with the invoice amount. For this team the fixed part is 6 × $25 + 2 × $125 = $400 a month on Claude Team billed monthly, or 6 × $20 + 2 × $100 = $320 a month billed annually, plus 8 × $19 = $152 for Copilot Business. A developer who pays for three tools personally adds up the same way: each plan’s price plus that plan’s measured overage. Whether every duplicated seat earns its place is a question for the cost per accepted change method, not for this table.

Prompts that turn billing exports into a checked estimate

Section titled “Prompts that turn billing exports into a checked estimate”

The agent does the extraction; the reconciliation rule makes the result checkable. Each prompt works in Claude Code, Codex, or Cursor in a folder that holds the exports.

For the security, data, and legal half of an enterprise purchase, use the vendor questionnaire alongside the third prompt.

How do you prove the cost estimate is right without reading every invoice line?

Section titled “How do you prove the cost estimate is right without reading every invoice line?”

You do not review the agent’s worksheet line by line. You check four things, and two people sign off:

  • Reconciliation. The worksheet total equals the invoice total per vendor, to the cent. The first prompt stops on any difference. This check enforces the same rule in CI or on your laptop: it exits non-zero on any non-numeric price (an unreplaced MEASURED or MISSING), on any vendor whose worksheet sum differs from its invoice total, and on any worksheet vendor with no invoice total. An overstated line at one vendor cannot hide an understated line at another:

    Terminal window
    # billing/invoice-totals.csv holds one row per vendor, from finance: vendor,total
    awk -F, 'FNR==1 { next } NR==FNR { inv[$1]=$2; next } { if ($4 !~ /^-?[0-9]+(\.[0-9]+)?$/) { print "non-numeric price on line " FNR ": " $0; bad=1 } else s[$2]+=$4*$5 } END { for (v in inv) { printf "%s worksheet %.2f invoice %.2f\n", v, s[v], inv[v]; if (sprintf("%.2f",s[v])!=sprintf("%.2f",inv[v])) bad=1 } for (v in s) if (!(v in inv)) { print "no invoice total for " v; bad=1 } exit bad }' billing/invoice-totals.csv billing/cost-worksheet.csv

  • Provenance. Every price row has a source and a checked_on date no older than 14 days when the purchase decision is made. A row with no source fails the check, whatever its value.

  • Seat roster. The count of named seats matches your identity provider’s list of people with access. Seats for people who left are the cheapest saving you will find.

  • Guardrails on variable spend. Every CI job that uses an API key has its own key and a cap (--max-budget-usd for claude -p; a workspace or key-level budget for Codex). The cost governance guide covers alerts below each cap.

Sign-off: the finance owner confirms the reconciliation against the invoices, and the tech lead confirms the seat roster and the CI keys. Neither has to read the other’s half.

Enterprise quotes: what to ask for

Section titled “Enterprise quotes: what to ask for”

Do not estimate an enterprise contract from public seat prices. Claude Enterprise, for example, adds usage at API rates on top of the seat. Ask every vendor to state in writing:

  • Platform and seat fees, minimum seats, and minimum term.
  • Included credits or token commitment, and the overage rate card with its date.
  • True-up rules when headcount or usage grows mid-term.
  • SSO, SCIM, data retention, data residency, and audit-log charges.
  • Charges for sandboxes, managed review, and automation runs.
  • Support tier and response times.

What goes wrong when you compare AI coding tool prices?

Section titled “What goes wrong when you compare AI coding tool prices?”
  • A price moves after you checked it. Vendors changed models, limits, and billing several times in 2026, including Copilot’s switch to AI credits on 2026-06-01. Recovery: keep a source and a date on every row, and re-open each source before you sign.
  • The same spend is counted twice. Claude Code’s /usage dollar figure on a Pro or Max seat, or a ccusage total, gets added to the subscription invoice. Recovery: invoices are the only source of cost; local trackers are for splitting cost between people.
  • “5x” is read as five times more messages. Plans are metered in usage windows and credits, not in messages. Recovery: measure one billing cycle of limit hits and blocked time before you move anyone up a tier.
  • Usage credits are switched on without a cap. Fast mode, Fable 5.1 and managed review all draw usage credits. Recovery: set a monthly spend limit before you enable credits (run /usage-credits, which opens Settings > Usage on Pro and Max and Admin settings > Usage on Team and Enterprise for members with billing access), and review weekly for the first month.
  • CI runs on someone’s personal seat or an untracked key. The spend appears on nobody’s budget until the invoice. Recovery: one API key per pipeline, a per-run cap, and the key’s owner named in the worksheet.
  • Shared consumer accounts to save seats. Recovery: one named seat per person; the saving is smaller than the audit gap it creates.
  • Old numbers are copied forward. Prices from search snippets or older articles are not verified. Recovery: read the vendor page the day you buy and put that date in the row.

Where to go next on AI tool cost

Section titled “Where to go next on AI tool cost”
The economics of agent-built software The one ROI method: cost per accepted change, with a worked example.
AI model comparison guide Per-token model prices, context windows, and defaults in one place.
What did the agents cost? /usage, ccusage, monitors, and gateway budgets for per-developer cost.
AI usage cost governance Budgets, alerts, and ownership for team-wide agent spend.
Size an AI plan from measured workload Choose a plan tier from completed-task cost and interruptions.
Buying AI coding tools The security, legal, and data questionnaire for an enterprise purchase.

Did this guide help you?

One tap, no sign-in.

not at all a lot

Frequently asked questions

Which Claude plans include Claude Code, and what do they cost?

On 2026-10-12 Claude Code is included in Claude Pro ($20 a month, or $17 a month billed annually), Max 5x ($100), Max 20x ($200), Team Standard ($25 a seat monthly, $20 annual), Team Premium ($125 monthly, $100 annual), and Enterprise ($20 a seat billed annually plus usage at API rates). The Free plan has no Claude Code. Since 2026-10-07 Max and Team plans also include monthly Claude API credits ($100 on Max 5x, $200 on Max 20x, $20 per Standard seat and $100 per Premium seat on Team); they do not apply to interactive Claude Code.

Is Codex sold as a separate subscription?

No. Codex is included in ChatGPT Plus, Pro, Business, Edu, and Enterprise plans. OpenAI's pricing page, read 2026-10-12, lists Free at $0, Go at $8, Plus at $20 a month, Pro at $100, $200 or $500 a month, Business at $20 per user per month billed annually ($25 billed monthly), and Enterprise and Edu by contacting sales. Pro plans currently have no five-hour limit, and the Ultrafast speed tier needs Pro at $500 or an eligible Enterprise or Edu plan. ChatGPT Pro at $100 gives 5x Plus usage on Codex and Pro at $200 gives 20x (secondary: TechCrunch and 9to5Mac, 2026-04-09).

What does GitHub Copilot cost for an agent workflow?

Copilot Pro is $10 a month, Pro+ $39, Max $100, Business $19 a seat, and Enterprise $39 a seat (GitHub docs, read 2026-10-12). Since 2026-06-01 Copilot bills usage in AI credits worth $0.01 each, with a base allotment per plan.

Where are Cursor's prices?

In the plans table below, read on cursor.com on 2026-10-12. Hobby is free, Start is ₹649 a month in India only, Pro is $20 a month, Pro+ $60, Ultra $200, Teams Standard and Premium are $40 and $120 per user per month ($32 and $96 billed annually), and Enterprise is custom. Cursor says usage beyond a plan continues at the same API rates, and Bugbot is billed by usage. No included-usage dollar amount per individual plan is verified here. Read cursor.com/pricing on the day you buy and record the date.

Where is the ROI calculation?

On the economics page, which defines cost per accepted change. This page keeps plan prices only, and per-token model prices live on the models hub.

Primary sources & references

All 577 guides Free resources: Compare tools Glossary By role

Written by

See what's new on the platform.

Privacy Policy Terms of Service Cookie settings llms.txt pricing.md RSS