Imagine an AI partner that not only autocompletes code but can handle multi-step development tasks for you. Enter GitHub Copilot’s new Agent Mode – an autonomous coding agent in VS Code that can analyze your entire workspace, make coordinated code changes across files, run tests, and even execute CLI commands as it iterates towards a goal.

In this forward-looking session, we’ll demonstrate Agent Mode by giving it a high-level task (for example, “add a new feature to this app”) and watching it break the work into steps, modify code, and self-correct errors in a loop. You’ll get to see how to configure and control the agent, including setting the right permissions and constraints so it remains a helpful servant, not a runaway script.

What You’ll Learn: Steps to enable Copilot’s agent mode in VS Code, and how to craft effective prompts that kick off complex tasks. We’ll peek under the hood at how the agent uses tools and context (and how you can supply custom tools or backends via the new Model Context Protocol, MCP, for even more power).

For hands-on learners, we’ll provide an example repository so you can try agent-driven development yourself. We’ll also discuss best practices and safety – e.g. reviewing AI-generated changes and gating what commands it can run – to use this automation confidently.

This is a glimpse into the future of development. By offloading routine and cross-cutting tasks to an AI agent, developers can dramatically boost productivity while maintaining full control over the final output. Whether you’re an early adopter looking to push the limits of AI in your workflow or just curious about the next evolution of Copilot, this session will inspire new ways to improve your developer experience and effectiveness.