AS Provider for Copilot

Use Anthropic, Google Gemini, OpenAI, and any compatible API in VS Code's GitHub Copilot Chat

English | 简体中文

Features

• Multi-Provider Support: Anthropic Messages, Google Gemini, OpenAI Responses, and OpenAI-compatible Chat Completions (compatible with DeepSeek, SiliconFlow, etc.)
• Automatic Model Discovery: Automatically enumerates available models via the provider's model list API — no need to manually enter every model ID
• Reasoning Effort Control: Supports five levels — low / medium / high / xhigh / max — for Anthropic and OpenAI reasoning models
• Image Input: Enables image context for models with vision capabilities
• Tool Calling: Enables agent tool-calling for models that support it
• Visual Configuration UI: Manage Provider Profiles, models, and global settings through a built-in interface or commands — no manual file editing required

Requirements

• VS Code >= 1.126.0
• GitHub Copilot extension (with chatProvider and languageModelSystem proposed APIs enabled)

Quick Start

1. Install AS Provider for Copilot from the VS Code Marketplace.
2. Press Ctrl+Shift+P and run AS Provider for Copilot: Open Configuration.
3. Add a Provider Profile, select the provider type, and enter the Base URL.
4. Enter the API Key for that profile.
5. Models will automatically appear in Copilot's model picker.

Visual Configuration UI

The extension provides a visual configuration interface to manage Provider Profiles, models, and global settings without editing any files manually.

Open the Configuration UI

Press Ctrl+Shift+P, search for and run AS Provider for Copilot: Open Configuration.

Multi-Provider Management

Each Provider Profile corresponds to a separate upstream API service. You can configure multiple profiles simultaneously and manage API Keys independently for each.

Reasoning Effort Control

For models that support reasoning (e.g. Claude, GPT), you can configure the reasoning effort level globally or override it per model using Model Overrides:

You can also use Model Overrides to set contextLength, toolCalling, and imageInput per model without affecting global settings.

Commands

Development

# Install dependencies
yarn install

# Full build (clean + WebView + extension)
yarn build

# Run tests
yarn test

# Lint
yarn lint

# Package .vsix
yarn build:vsce

src/
  core.ts          # Type definitions, provider options, message conversion
  discovery.ts     # Automatic model discovery
  extension.ts     # VS Code extension entry point
  storage.ts       # Configuration persistence
  views/
    configView.ts  # WebView controller
webview/           # Configuration UI frontend (Vue 3)
static/            # Built WebView static assets

Acknowledgements

• Vercel AI SDK — Underlying multi-provider inference driver
• VS Code Language Model Chat Provider API

Support & License

• Submit an Issue: https://github.com/Hoshino-Yumetsuki/as-compatible-copilot/issues
• License: MIT