Getting started with Roo Code in VS Code

Getting started with Roo Code in VS Code

Roo Code is a VS Code extension that can plan, edit files, and run terminal commands with a model you connect. We use it in the internship as a pair, not as a replacement for review.

Install

1. Open VS Code.
2. Install the Roo Code extension from the marketplace.
3. Reload the window.
4. Open the Roo panel and add a provider (OpenAI-compatible, Anthropic, or a local endpoint if you already run one).

Do not paste a production KILO_API_KEY or client key into a shared classroom machine. Students use their own keys or a classroom key that we rotate.

Modes we actually use

• Plan: describe the change, get a file list, reject anything that rewrites unrelated folders.
• Code: apply a small, named change (one route, one widget, one migration).
• Debug: paste a stack trace and the failing test, not the entire repository.

If the tool wants to "refactor the whole auth module" for a typo in a label, stop it. That is how student repos become unreviewable.

Rules we put in the repo

A short .roo or project instruction file helps:

• Prefer Bun and Elysia for new APIs in this repo.
• Do not add new dependencies without saying why.
• Do not commit secrets.
• Keep PRs under a few hundred lines when you can.

We still read the diff. Roo does not get a merge.

Local models

If you want code to stay on the machine, point Roo at a local OpenAI-compatible server (Ollama or similar). Quality is worse than a frontier API for large refactors. It is good enough for rename-and-explain work, which is most of what juniors need.

What this is not

Roo is not a substitute for the internship curriculum. Students who cannot explain the PR they "wrote" fail the review. The same is true on client projects: if nobody on the team can maintain the code, we did not ship.

For production AI features inside a product, that is custom AI development, not an editor extension.