AI Coding Assistant Pricing — Seats, Included Credits, and Overage

Pricing check — 1 October 2026: Copilot plan prices/allowances were rechecked against current GitHub Docs; Cursor plan structure and prices were rechecked against Cursor's live pricing page. Billing can vary by region, tax, and contract; confirm the checkout or organisation quote before purchase.

AI coding tools can combine a seat price with metered usage. GitHub says Copilot plans transitioned to GitHub AI Credits on 1 June 2026; individual plans have monthly allowances, while Business and Enterprise use pooled organization allowances. Some existing Pro/Pro+ annual subscribers can remain on legacy premium-request billing until that plan ends. Cursor's plans include model usage and let users continue with on-demand usage billed in arrears. These examples share a seat-plus-usage pattern, but their included limits, pooling and overage rules differ.

A mismatch between seat count and actual use can be easy to miss in the first month. Review seat utilization and credit burn after the initial trial and again at renewal; for example, check whether a small number of agent users accounts for most of a shared pool.

The three parts of a 2026 coding-assistant bill

1. The seat. Copilot Business is $19 per granted seat per month and Copilot Enterprise $39. Cursor Teams is $40 per user per month. These are current listed US prices before applicable taxes; enterprise contracts, billing currency and account terms can differ.

2. The included allowance. Current GitHub Docs list 1,900 AI credits per Business seat per month and 3,900 per Enterprise seat, pooled at the billing-entity level. Cursor says each plan includes a set amount of model usage. GitHub says paid-plan code completions and next-edit suggestions remain unlimited and do not consume AI Credits; Copilot model-using chat, CLI and agent features do.

3. The overage. After included credits run out, GitHub organization admins’ paid-usage policy and budgets determine whether usage continues or is blocked. Cursor says on-demand usage is billed in arrears after the included amount is consumed. In both systems, model and token use affect consumption, so estimate from representative work rather than seat count alone.

How the seat part still wastes money

Usage can vary widely across a team: some developers use an assistant daily, some occasionally, and others may stop after a trial. Measure use rather than inferring it from assigned licences, and reclaim seats that the team no longer needs.

The fix is unglamorous: review seat utilisation quarterly and reclaim dormant seats. With pooled credits there is a second reason to do it. A dormant seat still adds credits to the pool, which hides how fast the active users are burning through it.

How the usage part surprises you

Agent-style work — Claude Code, Codex, Copilot's agent mode, or Cursor's agents — can use substantially more model tokens than accepting occasional completions because it may involve long, multi-step tasks. Estimate it separately; one large migration can make a month unlike the previous one.

This is manageable, but only if you manage it:

• Set spending limits before rollout, not after the first surprise.
• Make the meter visible to the person delegating the task. Engineers make sensible decisions when they can see the cost, and wild ones when they cannot.
• Watch model choice. A chat model picker alone does not show total team consumption: Copilot credits cover multiple features, including CLI, cloud agent, Spaces and third-party coding agents. Reconcile the detailed usage report with the features your team actually used.

Forecasting credit burn in one afternoon

1. Export last month's usage by user and by model from the admin view.
2. Split users into three groups: completion-only (little credit burn), chat users (moderate burn), and agent users (heavy burn).
3. Price the pool. Copilot's allowances work out to a cent per credit at list price ($19 buys 1,900 credits). Multiply seats by the allowance, then compare with the burn of the agent group alone.
4. Decide where overage goes. Either adjust the org spending controls for named heavy users, or compare a separately metered agent workflow such as Claude Code API usage while keeping the seat for completions. For Codex, check the exact account and plan billing surface first.
5. Re-check after model launches. Model/API rate changes can alter the credit cost of a given workload; check the current GitHub model-pricing table rather than assuming every model has the same credit cost.

Where Claude Code fits

Claude Code is available through paid Claude plans with plan-specific usage limits, and through the Anthropic API with token-based billing. The limit/usage experience varies by plan and can change; do not assume a fixed weekly quota or treat API billing as a subscription seat. Claude Code limits after 14 September has the seat maths.

Which setup suits which team

If usage is broad and shallow, mostly code completion, a seat plan with unlimited completions can be predictable; compare the included AI-credit allowance with your team's actual chat and agent usage before calling it cheaper.

If usage is concentrated in a few engineers doing long agent tasks, compare shared-seat credits with a separately metered agent workflow. Avoid buying seats only to fund usage until a pilot shows the total cost and controls work for your team.

Many teams end up with both, and that is a legitimate outcome rather than a failure of discipline.

The costs that are not on the pricing page

Review time. An assistant that produces a large volume of plausible code shifts effort from writing to reviewing. That is usually a good trade, but it is not free, and it lands on your most senior people.

Switching cost. Adopting an AI-native editor means the team leaves an environment they have configured over years. That cost is real and one-off, and almost nobody models it. It is also the most common reason a promising trial quietly dies.

Before you sign

Run a two-week trial with a spending cap and measure actual usage rather than reported enthusiasm. Check which models the tool uses by default and what each costs in credits. Confirm what happens at renewal: promotional pools and introductory rates in this category have been short-lived. GitHub's current organization billing docs list standard allowances; confirm any account-specific or transition allowance in the billing view before forecasting.

To weigh cost against how autonomous you actually need the tool to be, the AI Coding Assistant Finder scores both together.

Sources (checked 2026-10-01)

• GitHub Docs — Plans for GitHub Copilot (Business $19 / 1,900 credits; Enterprise $39 / 3,900 credits; Pro, Pro+, Max allowances): https://docs.github.com/en/copilot/get-started/plans
• GitHub Docs — usage-based billing for organizations and enterprises (current standard credits, shared pool, budgets, overage controls and unlimited completions): https://docs.github.com/en/copilot/concepts/billing/organizations-and-enterprises/usage-based-billing
• GitHub Docs — usage-based billing for individuals (all plans, monthly credits, additional usage and annual-plan exceptions): https://docs.github.com/en/copilot/concepts/billing/usage-based-billing-for-individuals
• GitHub Blog — Copilot transition to usage-based billing on 2026-06-01: https://github.blog/news-insights/company-news/github-copilot-is-moving-to-usage-based-billing/
• Cursor — Pricing (Individual from $20/month; Teams $40/user/month; included usage, on-demand billed in arrears): https://cursor.com/pricing
• Anthropic — Claude Code product and available surfaces: https://www.claude.com/product/claude-code
• Anthropic — Claude API pricing: https://platform.claude.com/docs/en/about-claude/pricing
• OpenAI Codex — agentic coding product: https://openai.com/index/introducing-codex/