GitHub Copilot Enterprise Pricing and GitHub Copilot Business Pricing With AI Credit Cost

GitHub Copilot Enterprise costs $39 per user per month, and every assigned seat adds 3,900 GitHub AI Credits to a pool shared across your enterprise. One AI credit is worth $0.01, so the seat price is fully returned as usage. The catch is the prerequisite: Copilot Enterprise is sold only to organizations on GitHub Enterprise Cloud, which GitHub lists from $21 per user per month. A company that is not already on Enterprise Cloud pays at least $60 per developer for Copilot Enterprise.

Copilot Business costs $19 per user per month with 1,900 AI credits per user, and it works on any organization plan. On GitHub Team at $4 per user, a Business seat lands at $23 all in. That is the real price gap buyers should compare: $23 against $60 for a team on Team today, or $40 against $60 for a company already on Enterprise Cloud.

These are list prices before any overage. GitHub bills seats monthly for the number assigned at the end of each billing cycle. Seats added mid-cycle are charged pro rata and their included credits may be prorated too; seats removed mid-cycle are billed until the cycle ends. If the same person holds a seat in two organizations inside one enterprise, GitHub bills the enterprise once.

On credits alone, no. This is the single most useful number on the page. Enterprise costs $20 more per seat and includes 2,000 more AI credits per seat, and 2,000 credits at $0.01 is exactly $20. So the upgrade buys the same usage you could buy as overage on Business, except you prepay it for every seat whether that developer uses it or not.

Because both plans pool credits and both bill overage at the same $0.01, Business plus overage is always equal to or cheaper than Enterprise on usage. Here is the proof at 100 seats, Copilot seat cost plus overage only (add the GitHub plan on top):

The two lines meet once the average developer uses 3,900 credits a month and never cross. GitHub's own guidance says as much: it tells enterprises to treat the Business allowance as a baseline, and that the cheapest route for heavy agentic use is either upgrading or allowing additional usage with budget controls. The table shows which one wins on price.

So the Enterprise decision is a features decision. What the extra $20 buys besides credits, per GitHub: Copilot integrated into github.com as a chat interface across the platform, indexing of your organization's codebase for more tailored answers, access to fine-tuned custom private models for code completion, organization knowledge in GitHub Mobile, and priority or earlier access to new models and features. If none of those matters to a given organization, keep it on Business. GitHub lets an enterprise owner choose the plan per organization, so a pilot of Enterprise in one organization is the normal way to test it.

Every Copilot interaction except code completions is priced by tokens (input, cached input, cache writes and output) at the model's published per-million-token rate, then converted to AI credits at $0.01 each. A few of the rates GitHub lists for Copilot, per 1 million tokens:

Rates are hard to budget with, so we converted them into tasks. The token profiles below are our own illustrative assumptions, not GitHub figures, and they leave out cache write charges: a chat question at 20,000 input and 1,500 output tokens, an agent mode task at 300,000 fresh input, 1.2 million cached input and 40,000 output tokens, and a long cloud agent session at five times that.

Two conclusions hold whatever your exact token counts are. Model choice moves cost by roughly 30 times between the lightweight and the most powerful models, far more than the plan choice does. And a developer who runs a few frontier-model agent sessions a day will pass the 1,900 credit Business allowance within the first week, which is why the pool and the budgets matter more than the seat.

Each license adds its credits to one pool at the billing entity level. 100 Business seats make a single 190,000 credit pool, not 100 separate buckets, so a power user can draw on what lighter users leave. The pool resets to full at 00:00 UTC on the first day of each calendar month and unused credits are forfeited. Adding seats grows the pool immediately; removing seats shrinks it only from the next cycle.

The default is the part to act on. For organizations and enterprises, additional paid usage is enabled out of the box. If no one sets a budget, heavy agent use simply continues at $0.01 per credit. To forbid any spend above the included pool, an admin has to switch off the AI credits paid usage policy in AI Controls. When budgets run out, Copilot stops for that user until the next cycle; there is no longer an automatic fallback to a cheaper model the way there was under premium requests.

A sensible starting configuration for a 100-developer shop on Business: a universal user-level budget of about $40 (4,000 credits, roughly twice the per-seat allowance), individual overrides with an expiry date for the few engineers who run cloud agent all day, and an enterprise spending limit sized at 20% to 30% of the seat bill for the first two months while you learn the real consumption.

Four line items catch Copilot buyers after the contract is signed.

The last row explains most of the surprised September invoices. Teams that sized their habits on the summer promotion had 58% more usage per Business seat than they have now, and paid overage is on by default.

One more cost comparison belongs on the table. Copilot's meter charges for coding work you drive by hand. Some of that work is a repeated job that does not need a frontier model on a per-token meter at all: reviewing every pull request against your conventions, turning bug reports into labeled, routed tickets, or keeping API docs in step with the code. A ready-made agent that does one of those jobs runs on a flat monthly price with no credits to watch. The coding agents at the top of this page, Code Reviewer, Bug Triager and Docs Writer, connect to GitHub and are priced per month, not per token, so they are a predictable way to take the most repetitive volume off your Copilot pool.

If you are comparing coding agent costs across vendors, our Devin pricing breakdown covers Cognition's ACU model, our Claude Enterprise pricing page covers Claude Code seats for teams, and Cursor pricing covers Cursor's Teams seats and the token rate it adds to third-party models. For what changed from premium requests and how to convert an old budget, read GitHub Copilot premium requests pricing.