Opencode Go for Copilot

Opencode Go for Copilot adds OpenCode Go models to the VS Code Copilot model picker.

The extension is built for the native language model provider flow in VS Code 1.118+. Models stay visible in the picker before you add a key, most models use the OpenAI-compatible chat/completions path, MiniMax uses the Anthropic-style messages path, and reasoning output is preserved across tool loops where the upstream model docs require it.

What You Get

• OpenCode Go models under a single Opencode Go provider in Copilot Chat
• Silent model sync on startup and when opencodeGo.baseUrl changes
• Plain model names in the picker, without duplicated provider labels
• Vendor-specific thinking controls where the public docs support them
• Tool calling support through the current VS Code LM provider API

Supported Models

The shipped catalog includes the current OpenCode Go set:

• GLM-5
• GLM-5.1
• Kimi K2.5
• Kimi K2.6
• MiMo-V2-Pro
• MiMo-V2-Omni
• MiMo-V2.5-Pro
• MiMo-V2.5
• MiniMax M2.5
• MiniMax M2.7
• Qwen3.5 Plus
• Qwen3.6 Plus
• DeepSeek V4 Pro
• DeepSeek V4 Flash

Thinking Controls

Only models with public, documented request-side controls get picker options:

• DeepSeek V4 Pro / DeepSeek V4 Flash: Off, High, Max
• GLM-5 / GLM-5.1: On, Off
• Kimi K2.5 / Kimi K2.6: On, Off
• Qwen3.5 Plus / Qwen3.6 Plus: Auto, On, Off, plus optional Thinking Budget

MiniMax still preserves thinking history for multi-step tool use, but this extension does not expose a MiniMax on/off or effort picker because the OpenCode Go path does not document one clearly enough.

MiMo models stay on the default path for the same reason: public Xiaomi docs do not currently document stable request-side thinking controls for this integration.

Setup

1. Install the extension.
2. Run Opencode Go: Set API Key.
3. Open Copilot Chat and choose an Opencode Go model.

The API key is stored in VS Code secret storage. opencodeGo.apiKey still exists as a fallback setting, but the command is the cleaner path.

Commands

Notes

• The manual refresh command is intentionally gone. Startup sync remains silent.
• If you install another provider for the same vendor family, the Opencode Go entries still appear under Opencode Go, not under the other extension.
• Reasoning output is streamed when the backend returns it, and cached reasoning is replayed on later tool turns for providers that expect it.
• The extension targets VS Code 1.118+ and the current language model chat provider types.

Requirements

• VS Code 1.118+
• GitHub Copilot access for language model providers
• OpenCode Go subscription and API key

License

MIT