GitHub Copilot Enterprise pricing: $39 a seat, and what the extra $20 buys

The two organization plans

The corollary matters for a small team: two engineers on personal Copilot Pro+ at $39 each receive $140 of credits between them, while two Enterprise seats at $39 each receive $78. If you do not need the organization controls, the individual plans are strictly better value per dollar of inference, and by a wide margin.

What Enterprise adds for the extra $20

GitHub describes Enterprise as everything in Business plus priority access to new models and features, a larger monthly pool of AI credits, and enterprise-grade capabilities. Unpacked, the list is shorter than the price gap suggests.

Of those, codebase indexing on GitHub.com is the one that changes daily work, because it is what makes Copilot able to answer questions about a repository rather than about the file in front of it. If your engineers live in an IDE and never ask Copilot a repo-wide question, you are paying $20 a head for a capability nobody opens.

AI credits: the meter under both plans

GitHub replaced the old premium-request counter with AI credits. The mechanics are simple enough to model in a spreadsheet, and they are more forgiving than what they replaced.

• A credit is $0.01. Additional usage beyond the included allowance bills at $0.01 per AI credit.
• Credits pool at the billing entity. An enterprise with 100 Business licenses gets one shared pool of 190,000 credits, not 100 separate buckets, so a power user can draw far past their nominal share while light users offset it.
• Code completions stay unlimited on every paid plan and deduct no credits. Only chat, agents, code review, and similar features draw down.
• Budgets are the control, set at the user, cost center, and enterprise level.
• Each model carries a multiplier, so the credit cost of a message depends on which model answered it.

A worked check on a 200-engineer Enterprise deployment. Seats cost 200 times $39, or $7,800 a month. The pooled credit allowance is 780,000 credits, worth $7,800. If your measured agent usage is $60 per engineer per month, that is $12,000 of inference against $7,800 included, so you buy 420,000 credits at $0.01 and the real bill is $12,000, not $7,800. The seat line stopped being the budget the moment agents entered the workflow.

Seat management is where the money actually leaks

Copilot bills per assigned seat, not per active user, and the gap between those two numbers is the most reliable saving available in the whole product. The reclaim loop is fully automatable through the REST API.

gh api --paginate "/orgs/$ORG/copilot/billing/seats" \
  --jq '.seats[]
        | select(.last_activity_at == null
                 or (.last_activity_at | fromdateiso8601) < (now - 2592000))
        | .assignee.login' > dormant.txt

# review dormant.txt, then reclaim in one call
jq -Rn '{selected_usernames: [inputs]}' < dormant.txt \
  | gh api -X DELETE "/orgs/$ORG/copilot/billing/selected_users" --input -

Thirty dormant Enterprise seats is $1,170 a month, or $14,040 a year, for software nobody opened. Run the reclaim in the last week of every cycle, notify the affected people first so a returning engineer is not silently locked out, and keep assignment on teams rather than individuals so joiners and leavers follow the identity provider instead of a spreadsheet.

Copilot seats against an agent-first stack

The comparison people actually want is not Business against Enterprise. It is whether a Copilot seat is the right shape at all next to a terminal agent, because the two price completely differently.

Copilot’s honest advantage is that inline completions are unlimited and free of credits on every paid plan, which is a genuinely large amount of free value for the way most people still write code. Its honest disadvantage is that the agentic half, the part everyone is buying for in 2026, is metered at list price on top of a seat fee, so it converges on the cost of buying tokens directly while keeping a seat fee you cannot avoid.

Questions people ask

Keep reading

Continue through AI spend and governance with Claude Code for enterprise: policy, permissions, audit, and spend.