oneidea.ai
← Getting Started In AI  /  Online AI

Online AI

Build with the strongest models in the world — OpenAI, Anthropic and Llama — through simple cloud APIs. No powerful hardware needed. Three guides: your editor, then the provider setup, then wiring it all together.

Guide 1 — Install Visual Studio Code

VS Code is the free editor all three online providers plug into.

  1. Download VS Code from code.visualstudio.com and install for your OS.
  2. On Windows, tick ‘Add to PATH’ and the context-menu options during install.
  3. Open a project folder (File → Open Folder). Empty folder named ai-projects is fine.
  4. Sign in: VS Code has a built-in Settings Sync; a free GitHub or Microsoft account keeps your setup identical on every machine.

Guide 2 — Install extensions & set up each provider

Pick the provider(s) you want. You can install any combination — many developers use Codex or Claude for coding and Llama for cheap general tasks.

OpenAI — Codex

  1. Create an OpenAI account at platform.openai.com and add a small credit balance under Billing (a few dollars is plenty to start).
  2. Create an API key: platform.openai.com → API keys → Create new secret key. Copy it immediately — it is shown once only. Keep it secret: never commit it to code repositories.
  3. Install the Codex extension: in VS Code's Extensions view (Ctrl+Shift+X) search for OpenAI Codex and install it. There's also a standalone Codex CLI (npm install -g @openai/codex) if you prefer the terminal.
  4. Connect: open the extension panel and sign in with your OpenAI account (the extension handles the OAuth flow) or paste your API key when prompted.
  5. Test: select some code and ask the Codex panel to explain or improve it.

Anthropic — Claude Opus

  1. Create an Anthropic Console account at console.anthropic.com and add billing (small initial credit).
  2. Create an API key: Console → API Keys → Create Key. Copy it once and store it safely — treat API keys like passwords.
  3. Install the Claude Code extension: in VS Code, search Extensions for Claude Code (Anthropic's official extension) and install. Alternative: the standalone CLI (npm install -g @anthropic-ai/claude-code).
  4. Connect: sign in through the extension, or set your API key where the extension prompts for it. Pick the current Opus model in the model selector for maximum capability, or Sonnet for everyday speed.
  5. Test: open a project, ask Claude Code to run a small task end-to-end (e.g. “add an input validation function to this file”).

Llama API — Llama cloud models

  1. Get a Llama API key at llama-api.com (Meta's official Llama API) — create an account, accept the model terms, and generate an API key.
  2. Make it usable in VS Code: Llama's API speaks the OpenAI format, so it plugs into the same ecosystem. Install a versatile AI extension such as Kilo Code or Cline from the Extensions view, then add a custom OpenAI-compatible provider with base URL https://api.llama.com/v1 (check the exact URL in your Llama account dashboard — it is shown with your key) and your Llama API key.
  3. Pick a model: choose a current Llama model from the dashboard's model list (e.g. the latest Llama 4-family model) as the extension's model name.
  4. Test: ask it a question in the extension chat; responses come from Meta's models running in Meta's cloud.

Guide 3 — Good habits from day one

  1. Keys stay secret: never paste API keys into code you share, screenshots, or chat. Use environment variables — each provider's docs show the variable name (e.g. OPENAI_API_KEY, ANTHROPIC_API_KEY).
  2. Watch usage: platform.openai.com and console.anthropic.com both show live spend. Set a monthly limit on day one so a runaway script can't surprise you.
  3. Start cheap, scale up: frontier models cost more per request. Prototype with smaller/cheaper models, switch to Opus for the hard problems.
  4. Review before running: AI-generated code is a proposal, not a decision. Read diffs before accepting them — the extension shows you everything it wants to change.

You're done — what next?

You now have frontier AI models wired directly into your editor. Try: building a full small application in one conversation, generating tests for existing code, and asking for architecture advice on a real problem you have.

Curious about going fully private later? The Offline AI guide shows you how to run capable models on your own machine, free forever. And browse our UK AI directory to find UK companies building exactly this kind of technology.

Important disclaimer: the guides on this page are provided for information only. If you choose to install or use any software, extension or AI model described here, you do so entirely at your own risk and you are responsible for any and all issues that may arise — including, but not limited to, data loss, hardware problems, security or privacy incidents, software conflicts, unexpected costs, and results produced by AI systems. Always back up your files before installing software, and review everything an AI produces before relying on it.