VS Code's native AI: inline edits, chat, context providers

VS Code now ships AI features natively via GitHub Copilot integration — no separate IDE, no extension juggling. For teams that want AI without changing editors, this is the sweet spot.

What "native AI" means in VS Code

• Copilot Chat panel in the sidebar, always available.
• Inline chat via Cmd-I on any selection.
• Agent mode for multi-file tasks.
• Copilot edits for quick refactors.

If you have GitHub Copilot, you have all of this — no additional install.

The "don't move" value prop

You don't have to switch to a new IDE (Cursor, Windsurf) to get modern AI. Your extensions, shortcuts, debug configs, themes — all stay. For teams with existing VS Code investments (settings, workflows, team extensions), this matters.

Cursor and Windsurf are VS Code forks — similar but not identical. Upstream VS Code gives you the canonical experience.

Context providers

VS Code exposes context to Copilot via context providers:

• Current editor. What's visible in your current file.
• Workspace. Files in the open project, indexed.
• Terminal. Your recent terminal output.
• Problems. The Problems panel (TypeScript errors, lint, etc.).
• Extensions. Certain extensions (GitHub Actions, Azure, etc.) expose their own context.

Explicit references (#file, #selection, #problems, @workspace) pin these into the chat context.

The Copilot edit flow

For quick multi-file changes:

1. Cmd-Shift-I opens the edit panel.
2. Attach files (or skip; workspace auto-context is used).
3. Describe the change.
4. Review the proposed diffs across files.
5. Accept per-file or wholesale.

For most refactors, this is faster than Composer-style flows in forks because the integration with the rest of VS Code is tight.

Live share + AI

VS Code Live Share + Copilot is a genuinely useful pattern for pair programming:

• Two engineers in the same session.
• AI visible to both.
• Conversations flow into the codebase.

Underused; try it once and you'll find use cases.

Telemetry and privacy

VS Code Copilot has org-level controls:

• Data retention: prompt/completion pairs can be excluded from model training in enterprise plans.
• Content exclusions: certain file patterns blocked from AI.
• Per-team policy: which features available to whom.

Enterprise plans get strong enough guarantees to satisfy most policies.

Extensions that complement the native AI

• Error Lens — surfaces diagnostic messages inline; Copilot picks them up as context.
• Git Graph / GitLens — Copilot reads git context; these improve the surface.
• Todo Tree — centralizes TODOs; Copilot can tackle them systematically.

When to use VS Code AI vs. a fork

Note: Cursor and Windsurf ship AI features earlier; VS Code native catches up within months. If the cutting edge matters, fork. If stability matters, stay.

What to customize

• Copilot custom instructions in Settings.
• Keybindings for Chat, Inline Chat, Edit — pick ones you'll remember.
• Model preferences per task (Chat supports model selection).
• Excluded paths for sensitive code.