How to Use Grok for Coding in VS Code (Continue)

In short: create an xAI API key, install Continue (Continue.continue), set provider: xai with a current Grok model id from the xAI console, reload config, then chat. Model id strings change — copy them from the console, do not invent them.

Why Grok via Continue

You may already like Grok for reasoning, or your team standardised on xAI. Continue lets you keep VS Code and swap models without learning a new editor.

What you need

• xAI API key: from the xAI console. Store it as a secret, not in Git.
• Continue: VS Code extension ID Continue.continue (also works in many VS Code forks).
• config.yaml: usually ~/.continue/config.yaml (Continue gear → open config).
• Model id: exact string from xAI docs/console (example ids in Continue docs have looked like grok-4-1-fast-non-reasoning — re-check before you paste).

How it connects

Install steps

1. Create an API key in the xAI console. Copy it once; you will not see it again.
2. In VS Code: Ctrl+Shift+X → search Continue → install. Or: code --install-extension Continue.continue.
3. Open the Continue panel from the Activity Bar.
4. Open config (gear / “Open config.yaml”).
5. Add an xAI model block (sample below). Prefer Continue secrets / env substitution over pasting the raw key into a committed file.
6. Click Reload config. Select the Grok model in Continue’s model picker.

Configuration

name: Grok Coding
version: 0.0.1
schema: v1
models:
  - name: Grok (chat)
    provider: xai
    model: grok-4-1-fast-non-reasoning
    apiKey: ${{ secrets.XAI_API_KEY }}
    roles:
      - chat
      - edit
      - apply

1. provider must be xai (Continue’s xAI provider id).
2. roles: start with chat, edit, apply. Add autocomplete only if you pick a model Continue documents for that role.
3. Agent / tools: needs a model that supports tool use; if Agent fails, stay on Chat/Edit.
4. Same Continue install can also list Ollama or DeepSeek models — pick per task.

Settings tips

1. Keep Continue’s context small: reference one file with @ instead of the whole repo.
2. Turn off autocomplete if two AI completions fight (e.g. Copilot + Continue).
3. Check Continue docs → Model Providers → xAI when model names change.

First task (5 minutes)

1. Paste the prompt in Continue Chat.
2. Apply the edit to a new file only if the diff is small and clear.
3. Write or ask for a tiny pytest; run pytest -q yourself.
4. Reject any request to read .env.

Quick quiz

Save tokens and money

Grok API calls cost money. Use short prompts, one file at a time. Use local Ollama for practice drills; save Grok for harder coding.

Save these cards