BYOK for VSCode Copilot Chat

A BYOK extension for VSCode Copilot Chat.

Bring third-party model providers into VS Code Copilot natively. The extension supports both OpenAI-compatible and Anthropic-compatible services.

Features

• Supports two model providers:
  • OpenAI-compatible
  • Anthropic-compatible
• Supports custom API keys and endpoints
• Supports custom model lists with capabilities, context limits, and tool calling / vision / thinking flags
• Feels and behaves like native Copilot as closely as possible

How to Use

Add your model configuration:

• Open model picker -> Add Model -> OpenAI-compatible/Anthropic-compatible -> Follow Instruction

Then choose your custom model from the model picker:

You can start chatting right away after that. 😊

Configuration Example

{
	"name": "example",
	"vendor": "byok-anthropic",
	"apiKey": "${input:chat.lm.secret.-3b3a9fe3}",
	"endpoint": "https://api.example.org/",
	"models": [
		{
			"id": "claude-opus-4-6",
			"name": "Claude Opus 4.6",
			"vision": true,
			"toolCalling": true,
			"thinking": true,
			"adaptiveThinking": true,
			"supportsReasoningEffort": [
				"low",
				"medium",
				"high",
				"max"
			],
			"maxInputTokens": 872000,
			"maxOutputTokens": 128000
		}
	]
}

Field descriptions for models:

• id: The model identifier, such as claude-opus-4-7 or gpt-5.5
• name: The display name shown in the model picker
• vision: Whether the model supports vision input
• toolCalling: Whether the model supports tool calling
• thinking: Whether the model supports thinking mode

Other fields are only needed when you are defining a custom model.

• adaptiveThinking: Whether the model supports adaptive thinking
• supportsReasoningEffort: Supported reasoning effort levels, if applicable
• maxInputTokens: Maximum number of input tokens
• maxOutputTokens: Maximum number of output tokens

Acknowledgements

This extension was inspired by the DeepSeek V4 for Copilot Chat extension: https://marketplace.visualstudio.com/items?itemName=Vizards.deepseek-v4-for-copilot