How to use Claude Code with any model (Claude, GPT, DeepSeek…)

Claude Code is one of the best coding agents available, and it doesn't care where its tokens come from — it only needs an Anthropic-compatible endpoint. That means you can run it through Apixo and use any model we offer.

Step 1 — get your key and base URL

Sign in, open the dashboard and copy two values:

• your API key (sk-…)
• the base URL — for Claude Code, use it without /v1

Step 2 — configure Claude Code

The quickest way is environment variables:

export ANTHROPIC_BASE_URL=https://YOUR-BASE-URL
export ANTHROPIC_AUTH_TOKEN=sk-your-key
export ANTHROPIC_MODEL=claude-sonnet-5
claude

To make it permanent, put the same values in ~/.claude/settings.json:

{
  "env": {
    "ANTHROPIC_BASE_URL": "https://YOUR-BASE-URL",
    "ANTHROPIC_AUTH_TOKEN": "sk-your-key",
    "ANTHROPIC_MODEL": "claude-sonnet-5"
  }
}

Or skip the typing: Dashboard → Setup → Claude Code gives you this file with your key already inside.

Step 3 — switch models mid-session

Inside Claude Code, run:

/model gpt-6-luna

Any ID from the models page works. A common pattern:

• a strong model for planning and hard refactors,
• a fast, cheap model for routine edits and tests.

Why it stays cheap

Claude Code re-sends a large, mostly unchanged context on every turn. Those repeated tokens are served from the prompt cache and billed at the cache price — often around a tenth of normal input. Your usage page shows exactly how much caching saved you.

Troubleshooting