GitHub Makes Skills and MCP Context Generally Available in Copilot Code Review
Copilot reviews can now use repository skills and read-only MCP context across paid individual and enterprise plans.
GitHub made Copilot code-review support for agent skills and Model Context Protocol servers generally available on July 29, extending the feature to Copilot Pro, Pro+, Business and Enterprise users. The change lets repository and organization context participate directly in automated pull-request reviews, while GitHub says MCP tool calls made by the reviewer are restricted to read-only access.
What changed
Agent skills can now supply a team’s internal tools and coding standards during a Copilot review. GitHub’s setup uses a skill-specific directory under `.github/skills` with a `SKILL.md` file containing the relevant context and instructions. That gives teams a repository-native way to make review behavior reflect their own conventions instead of relying only on a generic review prompt.
MCP connections add another context path. GitHub says the reviewer can retrieve information from third-party systems such as issue trackers, documentation services and service catalogs. Existing MCP configurations for Copilot cloud agent automatically carry over to code review, and GitHub and Playwright MCP connections are enabled by default. GitHub also added attribution to comments produced with skill or MCP context, so reviewers can see when those inputs shaped a suggestion.
Teams that configured skills or MCP servers during the public preview do not need to change their setup. For new configurations, GitHub directs administrators to add MCP settings in repository Copilot controls and store authentication tokens under the Agents area of repository secrets and variables.
Why it matters
The practical shift is from a reviewer that sees primarily the proposed code and repository context to one that can consult organization-specific standards and connected systems. That can make an automated review more relevant to a team’s operating environment, especially when the correct decision depends on internal documentation or issue context. Attribution also gives human reviewers a visible clue about which external context contributed to a comment.
The read-only limit constrains what MCP tools can do during review, but it does not establish the quality of the resulting advice. GitHub’s announcement does not provide independent measurements of review accuracy, false-positive rates or security outcomes. Teams still need to inspect generated comments and govern the connected data and credentials.
Status
Confirmed. GitHub’s official changelog documents general availability, plan coverage, configuration, read-only MCP calls and comment attribution. Internal confidence is medium because the evidence is one vendor announcement without independent performance testing.
Sources
Update note: Last reviewed 2026-07-30. We will revise this post if GitHub changes availability, defaults or administration guidance.
Sources
- GitHub Changelog — official
Drafted with AI assistance from source briefs; reviewed for citation completeness and label accuracy.