GitHub Copilot code review comes to Azure Repos, billed per review

GitHub Copilot's automated code review is now open to every Azure DevOps customer, not just an early sign-up list. Each review is billed by token consumption, converted into GitHub AI credits at $0.01 each. The charge does not show up in Azure Cost Management until 48 hours later, according to Microsoft's own documentation.

How the billing actually works

There is no flat per-seat price during this preview. Cost depends on pull request size and the review effort level you choose, Low, Medium, or High. Microsoft's documentation states plainly that "higher effort levels generally analyze more context and consume more tokens" and therefore cost more.

That cost lands as GitHub AI credits, a unit Microsoft prices at $0.01 apiece. The charge is tied to your linked Azure subscription and tagged at the project level, so it can be traced back to a specific team's usage. It just will not appear until two days after the fact.

The limits that actually cap runaway spend

Budgets exist, but they are a warning system, not a brake. Microsoft's own words: "Budgets only notify you. They don't stop reviews or change any resources." A team that ignores the alert keeps generating billable reviews.

What actually throttles usage is concurrency, not cost. Azure Repos caps reviews at five running at once per organization, two per user, and one per pull request, per InfoQ's reporting confirmed by Microsoft's documentation. There are also hard size limits: repositories over 10 GB, or pull requests touching more than 100 files, are excluded.

What is actually new here

The core feature is not new. GitHub shipped it in technical preview back in June, then public preview in August, according to the Azure DevOps Blog. What changed this week, per InfoQ, is the removal of the sign-up gate. Also new: full disclosure of exactly how the billing and limits work, details that were not fully public before now.

Copilot's review itself stays advisory. It leaves comments but does not approve, block, or merge a pull request on its own; a human still owns that decision.

What this means for developers

Turn this on carefully, not by default. It sits off until enabled at the organization, project, or repository level, so you control the blast radius before flipping the switch.

Set your effort level deliberately. Low effort costs less per review but analyzes less context. A large or high-risk pull request may be worth the extra cost of a High-effort pass; routine changes can stay on Low.

Do not rely on a budget alert to protect you from surprise cost. It only emails you after the fact; it never blocks a review from running. If cost control matters to your team, track usage yourself against the concurrency caps rather than assuming the platform will stop you.

Expect the reporting lag when you check your bill. A review that ran this morning will not show up in Cost Management until the day after tomorrow, so don't panic if today's usage looks lower than you expected.