AI in VS Code Documentation
Microsoft's official guide to AI in Visual Studio Code: inline suggestions, chat, agents and the Plan agent, custom instructions, prompt files, permission levels, checkpoints and security. The most accurate source for how GitHub Copilot actually behaves inside VS Code, including every keyboard shortcut and slash command.
Free
VS Code Docs · Sept 2026
Guide
AI & Machine Learning
Microsoft
Reference, used alongside daily work
All
Free
What it covers
The VS Code documentation has a full section on building with AI in the editor. It is where Microsoft documents how GitHub Copilot and other agents work inside VS Code:
- Inline suggestions: ghost text, accepting the next word with Cmd+Right or Ctrl+Right, next edit suggestions and snoozing suggestions.
- Chat and agents: the Chat view, inline chat, the Plan agent, local and cloud agent sessions, and the Agents window.
- Customising:
.github/copilot-instructions.md,.instructions.mdfiles withapplyTopatterns, AGENTS.md, prompt files in.github/prompts, skills, custom agents, hooks and MCP servers. - Staying in control: permission levels (Manual, Assisted and Allow all), checkpoints, reviewing changes, agent terminal sandboxing and a security baseline.
- Reference: a features cheat sheet with the essential keyboard shortcuts and every chat slash command.
Is it worth reading?
If you use Copilot in VS Code, yes. The GitHub docs explain plans and GitHub.com features, but the editor experience, including shortcuts such as Ctrl+Alt+I for the Chat view and Ctrl+I for inline chat on Windows and Linux, is documented here. Two pages are worth reading in full early: the features cheat sheet, which is short, and the security page, which explains what auto-approval really means and why a Git worktree is not a security boundary.
How long it takes
The cheat sheet takes ten minutes. Custom instructions and prompt files take half an hour including trying them. The security and permissions pages take another half hour. After that, use it as a reference.
Why it matters for learning well
The docs are refreshingly direct about limits. They say that stopping a request does not undo actions already taken, that checkpoints do not reverse changes to external services, and that model-based approval can make mistakes. Those are exactly the habits the Sefism roadmap builds: review diffs, keep Git as your safety net and grant autonomy carefully.
How it fits the Coding with AI tools roadmap
The Copilot edition of the roadmap uses this guide for custom instructions and scoped rules in the setup stage, prompt files and the Plan agent in the prompting stage, fixing failing tests in the debugging stage, and permission levels in the agents stage.
Common questions
Is it only about Copilot? Mostly, but VS Code now supports other agent harnesses too, and the docs cover them.
Is it free? Yes, and VS Code itself is free.
Ready to start?
Opens on Microsoft in a new tab.
Stuck on something specific?
Writing only gets you so far. If you want an answer to your situation rather than the general case, book a session and we will work through it together. Sessions are free for approved Sefism members, and a few slots open each week.
Follow along
New writing, resources and project ideas land here first.