VS Code 1.111 launches Autopilot: an AI agent that writes, tests, and deploys on its own

VS Code 1.111 shipped on March 9, 2026, and it's not another release of minor fixes. Every new feature in this version is tied to AI. The headline addition is Autopilot mode, in which the agent writes, tests, and deploys code on its own, without stopping or asking for confirmation until the entire task is done.

What is Autopilot mode

Autopilot is not a souped-up autocomplete. It's a full AI agent that runs in a loop: it takes a task, plans the steps itself, carries them out, and corrects the result when needed.

The key difference is that the agent doesn't wait for the developer at every step. It approves its own tool calls, retries on its own when something fails, and answers the intermediate questions that can come up along the way.

Microsoft lead engineer Kai Maetzel put it bluntly: the agent shouldn't "sit idle waiting for a response." The goal is to remove every point where the process slows down because it has to wait for a human.

This is a fundamentally different model from anything that came before. GitHub Copilot offered autocomplete suggestions, and a human picked one. Copilot Chat explained code and generated functions on request, but the decisions stayed with the developer. Autopilot is the first to take the human out of the decision loop while a task is running.

What the agent does without the developer

Autopilot covers the whole typical cycle, from writing code to deployment:

• generates and edits code from a task description
• runs tests and interprets the results
• fixes compilation errors and linter warnings
• uses the terminal and runs build commands
• deploys changes through a defined pipeline

The developer sets the task and gets a finished result. What happens between those two points is up to the agent.

Is this a good thing

The author of the original piece has been writing code since the early 2000s. He watched Eclipse slow to a crawl, IntelliJ take over the Java world, and VS Code become the default editor for an entire generation.

In his view, Autopilot isn't just another feature. It changes the very nature of the relationship between the developer and the tool.

"This is different.

This changes the relationship between the developer and the tool itself."

The risks are obvious. An agent that approves its own actions moves fast, but when something goes wrong, the developer finds out afterward, not while it's happening. A deployed bug, an overwritten config, a command run in prod instead of dev: all of this becomes possible once the human is removed from the chain.

Autopilot is an explicit bet on speed, and each team will have to decide whether that bet is worth the loss of control.

What it means

VS Code has gone from an editor with autocomplete to an autonomous task executor, and that shift took less than two years of active AI feature development.

For teams, Autopilot offers real time savings on routine cycles. But it will only work in an environment with explicit guardrails and clear rules. Without them, autopilot mode doesn't speed things up; it becomes a source of risks that are harder to catch than ordinary bugs.