Back to rankings

GitHub Copilot Code Review: Agent Skills & MCP Reach General Availability

GitHub CopilotView original changelog

GitHub Copilot code review now supports agent skills and MCP servers in general availability for all Copilot Pro, Pro+, Business, and Enterprise users, graduating both capabilities out of public preview. Agent skills let teams drop a SKILL.md file under .github/skills so Copilot can apply internal tools and coding standards during a review, while MCP server connections pull in read-only context from third-party platforms such as issue trackers and documentation systems. Any MCP configuration already set up for the Copilot cloud agent now applies automatically to code review, with the GitHub and Playwright MCP servers turned on by default. Review comments generated with skill or MCP context now carry attribution labels, so teams can see exactly which extension informed a given suggestion.

Key Takeaways

  • Agent skills and MCP support for Copilot code review are now generally available for Pro, Pro+, Business, and Enterprise plans, ending the public preview period.
  • Teams define custom review behavior through a SKILL.md file placed in a .github/skills subdirectory, tailoring Copilot's analysis to their own standards.
  • MCP server connections bring in outside context — issue trackers, documentation systems, service catalogs — directly into the review process.
  • All MCP tool calls made during a review are strictly read-only, limiting the blast radius of any external integration.
  • Existing Copilot cloud agent MCP configurations carry over automatically to code review, with the GitHub and Playwright MCP servers enabled by default.
  • New attribution labels on review comments let teams distinguish baseline Copilot feedback from suggestions shaped by a custom skill or MCP source.

Overview

GitHub Copilot code review has graduated its support for agent skills and Model Context Protocol (MCP) servers from public preview to general availability, rolling the capability out to every Copilot Pro, Pro+, Business, and Enterprise subscriber. Together, the two features let a code review pull in context that goes well beyond the diff itself — a team's internal tooling, coding conventions, and third-party knowledge sources can now shape every automated review Copilot performs.

How Agent Skills Extend Reviews

Agent skills give Copilot code review a way to invoke a team's own tools and standards while it works through a pull request. Setting one up means creating a skill-specific subdirectory under .github/skills and adding a SKILL.md file describing the context and instructions Copilot should apply. Because these skills live in the repository itself, they travel with the codebase and can be scoped to a specific project, team, or organization rather than living as a one-off prompt.

How MCP Server Support Works

MCP server connections let code review pull live context from external systems developers already rely on, such as issue trackers, documentation platforms, and service catalogs. Every MCP tool call made during a code review is restricted to read-only access, and any MCP servers already configured for Copilot's cloud agent carry over automatically — no separate setup is required. The GitHub MCP server and the Playwright MCP server are both enabled by default, giving reviews an out-of-the-box way to cross-reference repository metadata and browser-based checks.

Attribution on Skill- and MCP-Generated Comments

A new attribution layer now marks review comments that were shaped by an agent skill or MCP context, making it clear when a suggestion came from one of these extensions rather than Copilot's baseline analysis. This gives reviewers a way to audit how much of the feedback on a given pull request is coming from custom, team-specific configuration versus the model's default behavior.

Getting Started

Teams that already configured agent skills or MCP servers during the public preview do not need to change anything — existing setups carry over unchanged now that the features are generally available. Teams setting this up for the first time can add MCP configuration under repository settings, storing any authentication tokens under the repository's secrets and variables settings, and can add agent skills by creating a .github/skills subdirectory with a SKILL.md file describing the desired review behavior.