Free AI Coding in VS Code

You do not need a subscription to code with AI. I present a free setup using an open source extension called Continue, and a free API key from Google AI Studio. Setup takes about five minutes.

What is a coding harness?

A language model on its own is just text in, text out. It cannot see your project. It cannot open a file, run a test, or apply an edit.

A coding harness is the software that wraps the model and gives it hands. It decides which files to send as context. It formats your question into a prompt. It reads the reply and turns it into an edit in your editor. Think of the model as an engine and the harness as the car around it.

Continue is the harness and it is free forever. Gemini is the model and Google gives away a free tier. Using both, you have a free AI coding assistant in VS Code.

Step 1: Install Continue

Continue is an open source harness, bought by Cursor, but still freely available. The source is on GitHub if you want to see how it works.

Open VS Code.
Press Ctrl + Shift + X to open Extensions. On a Mac use Cmd + Shift + X.
Search for Continue. The publisher is Continue.
Click Install. You can also grab it from the VS Code Marketplace.
A Continue icon appears in the sidebar. Click it to open the chat panel.

Continue will ask you to pick a model. Ignore that for now. Get your key first.

Step 2: Get a free Gemini API key

Google AI Studio hands out API keys at no cost. You need a Google account and nothing else.

Go to aistudio.google.com/api-keys.
Sign in with your Google account.
Click Create API key.
You might be asked to make a project first, give it an arbitrary name.
Copy the key once it's generated.

Treat this like a password Never commit an API key to git. Never paste it into a public repo, a screenshot, or a group chat. If a key leaks, delete it in AI Studio and create a new one.

Step 3: Paste the key into Continue

Open the Continue panel in VS Code. There is a field for API Keys. Paste your key under the appropriate provider. (Gemini if you are using the free Google key.)

Set the provider to Google Gemini if you are using the free Google key.
Paste your API key into the key field.
Click Connect. Continue writes the config for you.

Continue reads a YAML file at ~/.continue/config.yaml. On Windows that is %USERPROFILE%\.continue\config.yaml. It looks like this:

name: My Config
version: 0.0.1
schema: v1

models:
  - name: Gemini Flash
    provider: gemini
    model: gemini-3.7-flash
    apiKey: YOUR_KEY_HERE
    roles:
      - chat
      - edit
      - apply

Save the file. Continue picks up the change straight away.

Step 4: Select a Flash model

Google ships several Gemini models. Pick a Flash one.

Flash models are fast, cheap, and generous on the free tier. They are strong enough for coursework, debugging, and explaining code.
Pro models are smarter but burn your free quota in a handful of requests.

Choose the latest Flash in the dropdown. Today that is gemini-3.7-flash, though Google renames these often. If the name changed, take the newest one with Flash in it.

You are done That is the whole setup. Continue is now wired to a real model, and you are not paying for it.

Step 5: Actually use it

Continue has three modes.

Chat. Highlight code, press Ctrl + L, ask a question. Good for "why does this throw" and "explain this loop".
Inline edit. Highlight code, press Ctrl + I, describe the change. Continue rewrites it in place and shows a diff.
Autocomplete. Suggestions appear as you type. Press Tab to accept.

Start with questions rather than commands. Ask it to explain an error before you ask it to fix one. You learn more that way, and you catch it when it is wrong.

My Suggestions

Read every suggestion before you accept it. The model is confident even when it is wrong.
Never send secrets. No API keys, no passwords, no personal data in a prompt.
Free means you are the product, your prompts help train the AI. Do not paste anything confidential into the free tier.

When you outgrow the free tier

The free setup is great for learning. Paid tools pull ahead on long, multi-step work: refactoring across many files, running tests in a loop, working from a terminal. These are worth a look once you know the workflow.

Claude Code Anthropic's terminal and IDE agent. Strong at large refactors and multi-file work. Paid plan or API credits Codex OpenAI's coding agent for the CLI, IDE, and cloud tasks. Included with paid ChatGPT plans Cursor A fork of VS Code built around AI. Familiar if you already use VS Code. Free tier, then monthly GitHub Copilot Deep VS Code integration. "Free" for verified students through GitHub Education. Free for students

Continue also works with those paid models. Swap the provider and key, keep the same harness.