Commandcode Provider

Command Code provider for OpenCode, built on Command Code's official Provider API.

Community-built and not affiliated with or endorsed by Command Code.

• Uses the documented https://api.commandcode.ai/provider/v1 endpoints, not the internal CLI bridge.
• Fetches the live model catalog on every startup, so new models appear with no manual sync.
• Falls back to a local cache when the catalog request fails.
• Registers its own auth method so /connect works.

Install

Add the plugin to ~/.config/opencode/opencode.json (or opencode.jsonc):

{
  "$schema": "https://opencode.ai/config.json",
  "plugin": ["@yusuf.ameri/opencode-commandcode-provider"]
}

Then connect once:

/connect

Search for Command Code and paste your API key. The same key that authenticates the Command Code CLI works here.

Alternatively, provide the key with an environment variable:

export COMMANDCODE_API_KEY=your-key

Models

The plugin fetches GET /provider/v1/models and registers every model served over the OpenAI-compatible /chat/completions endpoint. Claude models are /messages-only and are intentionally excluded from this provider; configure them separately if your plan includes them.

Models are cached at $XDG_CACHE_HOME/opencode/commandcode-models.json (~/.cache/opencode/commandcode-models.json by default) and refreshed on every startup.

Credential lookup

The API key is resolved in this order:

1. COMMANDCODE_API_KEY
2. ~/.local/share/opencode/auth.json under commandcode
3. ~/.commandcode/auth.json (the Command Code CLI login)

Configuration

The plugin only fills in fields you have not set. To override anything, declare the provider yourself:

{
  "provider": {
    "commandcode": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "Command Code",
      "options": { "baseURL": "https://api.commandcode.ai/provider/v1" }
    }
  }
}

Development

bun install
bun run typecheck
bun test

License

MIT