OpenCode - appleai.codes Docs

OpenCode works best when you define a dedicated provider block and keep the routed model path explicit.

OS notes

macOS and Linux: Use your normal shell and keep the JSON config exactly as shown
Windows: Use the same JSON config and paste it through the Windows app or PowerShell-based workflow your install exposes
WSL: If OpenCode runs inside WSL, follow the Linux conventions from inside WSL

Example config

Add a provider block like this to your OpenCode config:

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "appleai": {
      "npm": "@ai-sdk/openai-compatible",
      "options": {
        "name": "AppleAI",
        "apiKey": "YOUR_APPLEAI_API_KEY",
        "baseURL": "https://api.appleai.codes/v1"
      },
      "models": {
        "claude-sonnet-4": {
          "name": "anthropic/claude-sonnet-4"
        },
        "gpt-5": {
          "name": "openai/gpt-5"
        }
      }
    }
  },
  "model": "appleai/claude-sonnet-4"
}

What to verify

The provider name and the model prefix line up
baseURL includes /v1
Each OpenCode model alias maps to the routed model you want upstream

If you rename the provider block, update the leading segment in the model field to match.