GitHub Copilot cloud agent turns issues into reviewable branches

GitHub's Copilot cloud agent documentation shows a coding-agent pattern that is less about an IDE chat session and more about delegating repository work into a hosted, auditable workflow. Copilot can research a repository, create an implementation plan, make code changes on a branch, iterate before a pull request, or open a pull request directly from GitHub issues, Copilot Chat, pull request comments, scheduled automations, and security campaigns. The agent runs in its own ephemeral GitHub Actions-powered development environment where it can explore code, install dependencies, run tests and linters, and surface logs. Repository owners can make that environment more deterministic with `.github/workflows/copilot-setup-steps.yml`, runner selection, permissions, services, snapshots, and a hard timeout of up to 59 minutes. GitHub's MCP configuration docs add the next operational boundary: repository-level MCP settings can expose selected tools to Copilot cloud agent and Copilot code review, but tool allowlists, secrets prefixed with `COPILOT_MCP_`, and the warning that configured tools can be used autonomously become governance requirements. For agent builders, the practical signal is to evaluate hosted coding agents by the whole issue-to-branch-to-PR contract: entry points, branch isolation, environment reproducibility, logs, tool permissions, human review, custom agents, hooks, skills, cost metrics, and whether every autonomous action remains inspectable through GitHub.