Microsoft has opened GitHub Copilot code review for Azure Repos to all Azure DevOps customers, removing the sign-up that gated it since June. The feature posts review comments on pull requests, bills per review through the linked Azure subscription, and reports spending two days after it happens.
The June announcement explained why the feature exists, in terms Microsoft rarely puts in writing. Dan Hellem and Andrew Brenner wrote that the company had spent several years encouraging customers to move repositories from Azure Repos to GitHub for its AI and agentic development experiences, then acknowledged the result:
While many customers are actively planning and executing migrations to GitHub, many others are not yet ready to move and continue to rely on Azure Repos for day-to-day development.
Migration complexity varies with an organization's size, customizations, compliance requirements, tooling, and industry constraints. Rather than wait for those customers to move, Microsoft is bringing the GitHub capability to them.
Customers read the release the same way. Responding to the announcement, commenter Marc Selman wrote:
Features like this have been available on GitHub for a long time already. For a lot of businesses and enterprises it's too much work and risk to simply migrate to GitHub. I'm glad they're finally arriving to DevOps.
That pattern extends beyond code review. Microsoft's Azure DevOps Remote MCP Server reached general availability this month, giving AI assistants a hosted endpoint into work items, pull requests, repositories and pipelines, though Claude, ChatGPT and Cursor cannot yet connect to it because Entra lacks the client registration mechanisms those clients require. Both releases serve the same population: teams that want agentic tooling in Azure DevOps without moving their repositories.
The August release adds onboarding controls at organization, project, and repository level, support for Managed DevOps Pools, custom instructions, and automatic reviews driven by branch policies. Custom instructions can be applied across an organization or project, set per repository, or targeted at specific paths.
Billing is where the release deserves a close reading. Each completed review consumes input, output, and cached tokens, converted into GitHub AI credits at one credit per cent. Higher review effort analyzes more context and consumes more, and pull request size, custom instructions, and model changes all affect the total. Charges appear in Azure Cost Management under the product GitHub Copilot for AzDO, 48 hours after a review completes.
Cost attribution improved with this release. Charges now carry Azure DevOps project tags, so teams can filter or group cost analysis by project instead of seeing an organization-level total. Because these are standard Azure tags, they work in cost analysis views, exports, and budgets.
The documentation is explicit about what a budget does, and does not do:
Budgets only notify you. They don't stop reviews or change any resources.
A team can set a threshold and receive an email within an hour of the next evaluation. Combined with the 48-hour reporting delay, an automatic review policy applied across a busy project would run for two days before its cost appeared anywhere.
Concurrency limits are worth checking before enabling automatic reviews widely: five concurrent reviews per organization, two per user, one per pull request. An organization with several teams reviewing every new pull request will queue.
The feature runs on Azure Pipelines infrastructure. By default, it uses the organization's default agent pool, which some organizations disable. Managed DevOps Pools are supported and must run the latest Ubuntu Server image; Windows images are not supported, and Microsoft advises against mixing Ubuntu and Windows agents in one pool. Self-hosted agents are not supported.
The review itself is advisory by design. Copilot always leaves a Comment review, never approves or requests changes, so its feedback does not satisfy required-reviewer policies and does not block a merge. It does not re-review after new commits unless asked, and it does not read replies or follow up.
Reliability during the preview has been uneven. A customer reported that reviews stopped after the plan-tool execution phase, remained idle for roughly 60 minutes, and were then automatically canceled. Hellem confirmed the behavior as a known issue with a fix rolling out that would take about a week to reach all organizations. Another customer could not find the automatic review setting where the documentation described it, and Hellem clarified that it sits at branch policy level.
The preview terminology has been inconsistent throughout. The June post described a limited public preview while running a sign-up it called a technical preview. The August post calls it public preview with no sign-up required. The documentation still carries a limited preview banner, with no SLA and limited support.
Requirements constrain where the feature applies. Repositories must be 10 GB or less, pull requests must contain 100 changed files or fewer with no merge conflicts, and TFVC is not supported. Rollout is gradual by region, and Microsoft expects two to three weeks or longer to reach all organizations.
Microsoft's own guidance is to enable the feature for one or two repositories, compare effort levels, and monitor daily usage before expanding. The cost of a review scales with pull request volume, and the reporting arrives two days later.