opencode-commandcode

An OpenCode plugin that registers Command Code as a provider. It supports API-key login and live model discovery against the Command Code Provider API.

Install

Add the plugin to your opencode.json:

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

Log in

Run the auth flow and paste your API key when prompted:

opencode auth login

Select Command Code from the provider list. Your key is stored in OpenCode's credential store (auth.json) — the same key used by the Command Code CLI and API.

Alternatively, skip auth login by exporting the key:

export CMD_API_KEY=...

Use

Models are discovered automatically from GET https://api.commandcode.ai/provider/v1/models at startup:

opencode run --model commandcode/deepseek/deepseek-v4-flash

The full model list (Claude, GPT, Gemini, DeepSeek, Qwen, Kimi, and more) is available under the commandcode/ prefix. Model names and context limits come from the live endpoint; output limits and vision/reasoning capabilities come from a bundled override table.

How it works

The plugin exposes two hooks:

• auth (type: "api") — registers the commandcode provider with OpenCode's login flow so opencode auth login can store an API key.
• config — injects config.provider.commandcode with @ai-sdk/openai-compatible pointed at https://api.commandcode.ai/provider/v1, plus the live model list (with a static fallback when the endpoint is unreachable).

If the /models endpoint cannot be reached at startup, a small fallback list is used so the provider remains usable. The list is never left empty.

Development

npm install
npm run build

License

MIT