
GitHub Copilot's autonomous agent mode hits GA
Copilot Agent Mode is generally available on VS Code and JetBrains. The agent picks files to edit, runs terminal commands, and iterates on errors without manual intervention. The Coding Agent for PRs ships in parallel.
GitHub announced general availability of Copilot Agent Mode on March 12, available across VS Code and JetBrains IDEs [GitHub press release].
── What shipped ──
Agent Mode is autonomous multi-step coding. The agent decides which files to edit, executes terminal commands, runs tests, and iterates on its own errors — no per-step user prompting required.
Coding Agent for PRs. Assign a GitHub issue to Copilot. The agent works in the background, writes code, runs tests, and opens a PR for review. This shipped alongside Agent Mode GA.
Agentic code review. Copilot's code reviewer now gathers full project context before suggesting changes and can pass those suggestions directly to the coding agent for fix-PR generation [GitHub docs].
Workspace context. Agent Mode automatically picks up file structure, open files, and project configuration without manual file tagging.
── Why it matters ──
The honest read: this is GitHub catching up to Cursor on autonomous agents inside the IDE, and to Cognition's Devin on issue-to-PR automation. The catch-up is significant because GitHub's distribution dwarfs both — Copilot has been bundled into nearly every enterprise GitHub Enterprise deal for years.
For most developers, this is the first time autonomous-agent coding is available in the IDE they already use, billed against the seat their employer already pays for. The friction-to-adoption number drops to near zero.
For Cursor and the agent-IDE startups: the moat is now product velocity, not access. Copilot has reach; Cursor has model-routing flexibility, faster shipping cadence, and a sharper product. Whether that is enough to defend a $50B valuation is the open question.
── Editor's take ──
Agent Mode at GA is the moment "agentic coding" stops being early-adopter territory. Expect every enterprise codebase to see at least one Copilot-authored PR within the next quarter. The interesting metric to track is rejection rate. If it's below 30%, this changes how teams operate. If it's above 60%, this is another autocomplete generation. Six months will tell.
// newsletter_offline · provider_not_configured