What changed
GitHub said on July 29 that Copilot code review support for agent skills and MCP is now generally available. That moves these capabilities from preview into the default path for paid Copilot teams that want review comments grounded in repo-specific instructions and connected tools.
In practice, this means a review can now use repository skills defined in .github/skills plus MCP-backed context from the systems a team already uses. GitHub's announcement also says MCP calls in code review are limited to read-only access, and that existing cloud-agent MCP setups carry over into code review.
Why it matters
This is a meaningful product step because code review stops being just a generic LLM pass over a diff. Teams can encode their own review standards in agent skills, and they can connect Copilot to external context through MCP server integrations.
That changes the shape of AI review from "does the model notice something" to "does the review have the same context your team uses to make decisions." For product and engineering teams, that is the difference between novelty and a workflow you can actually trust more often.
What builders should take away
The practical takeaway is to treat code review as another programmable agent surface. If your team already writes repeatable review guidance, move it into SKILL.md files. If your reviewers constantly need Jira tickets, docs, service catalogs, or test-system context, wire that in through MCP where read-only access is enough.
GitHub also added attribution on comments generated with skills or MCP context, which matters operationally because reviewers can see when extra instructions or external tools influenced a suggestion. And for org-managed Copilot deployments, GitHub's MCP docs note that MCP access can be governed by policy, which gives platform teams a cleaner control point as this spreads.
Bottom line
GitHub is making AI code review feel less like a standalone assistant and more like a programmable repository workflow. For teams already experimenting with Copilot customization, this GA release makes the review layer one of the most practical places to productize those investments.