Did GitHub add agent skills and MCP support to Copilot code review?

Did GitHub add agent skills and MCP support to Copilot code review Yes. GitHub added agent skills and MCP support to Copilot code review, and GitHub documents both features in its changelog and product docs. The change is real, current, and tied to review workflows, not a generic Copilot feature.

GitHub’s changelog says Copilot code review support for agent skills and MCP servers is now generally available for Copilot Pro, Pro+, Business, and Enterprise users. The same post says these capabilities let teams bring internal tools, coding standards, and external context into code review.

GitHub’s code review docs say Copilot code review can use repository-level agent skills and MCP servers when they are relevant to the review. The docs also say repository MCP configuration applies to both Copilot cloud agent and Copilot code review, which is the part people often miss.

Agent skills are not a UI toggle that magically changes every review. GitHub says they live under .github/skills, and Copilot code review can read relevant skills from the head branch, the branch with the pull request changes. That matters because you can test a skill change in the same pull request without merging it first.

MCP support is also not just “Copilot can browse the internet.” GitHub says Copilot code review can use MCP servers to pull context from third-party systems such as issue trackers, documentation systems, service catalogs, and incident tools. GitHub also says those MCP tool calls are read-only for code review.

The inconvenient part is that Copilot will not use every skill or every server on every review. GitHub says it is more likely to use them when the repository or pull request gives clear signals, such as review-focused skill directory names, custom instructions that reference MCP context, or pull request descriptions that include identifiers tied to configured MCP servers. A vague setup can sit there and do nothing useful.

Another detail people get wrong is scope. GitHub says the GitHub MCP server and Playwright MCP server are enabled by default for code review, but repository settings control whether Copilot can use MCP tools while reviewing pull requests. If you disable that setting, MCP stays available for Copilot cloud agent but not for code review.

The short version is simple: GitHub did add both agent skills and MCP support to Copilot code review, and the feature is documented as part of the review product itself. If you are setting it up, start with the GitHub docs for code review, then check your repository skills folder and MCP settings in the same branch you use for the pull request.

If you are organizing this kind of AI-assisted workflow across a team, DevConnect keeps the practical side separate from the product marketing side: https://devconnectplatform.com. The useful test is still the same, whether the review agent can see the right context and whether your repository rules make that context actionable.

What this does not mean is that every review becomes better automatically. Skills need clear naming and clear instructions, MCP needs a configured server with the right context, and the pull request itself has to point the agent toward the right signal. Without those pieces, Copilot review still runs, but it has less to work with.

For a team that wants to verify the setup, the clean check is to add a review-focused skill under .github/skills, connect one MCP server in repository settings, and open a pull request that includes the identifiers or files the skill expects. If Copilot uses them, you will see attribution in the review comments. If it does not, the setup needs clearer signals, not more hope.

In practice, this is a workflow feature, not a slogan. GitHub has added the plumbing, documented where it lives, and made it available in code review. The remaining work is repository design: naming, instructions, settings, and the shape of the pull request itself.