Omni Chat Provider

Use OpenAI, Anthropic, Gemini, Ollama, and OpenAI-compatible backends inside VS Code Copilot Chat through one extension.

The extension contributes a single vendor, OmniChat, to Copilot's language model system. Each model group in Chat: Manage Language Models points at one configured providerId, so removing a group only removes the Copilot-side mount. Your own omnichat.providers, omnichat.models, and stored API keys stay intact.

Features

Native adapters for openai, openai-responses, anthropic, gemini, and ollama
Provider-level connection settings with model-level overrides
Per-provider API keys stored in VS Code Secret Storage
Native Copilot model management via Chat: Manage Language Models
Stable internal model routing with clean picker labels
Retry, delay, custom headers, and system-prompt interception
Optional model variants via configId
Commit message generation command

How It Works

There are three layers:

omnichat.providers Defines backend connections such as baseUrl, apiMode, and shared headers.
omnichat.models Defines the actual models shown in Copilot, each mapped to a provider via provider or owned_by.
Chat: Manage Language Models Creates Copilot model groups that attach one providerId to the OmniChat vendor.

Deleting a Copilot group does not delete OmniChat settings or OmniChat secrets.

Requirements

VS Code 1.104+
GitHub Copilot Chat installed

Setup

1. Install the extension

Install the VSIX or install from Open VSX after publishing.

2. Define providers

Add provider backends in settings.json:

"omnichat.providers": [
  {
    "id": "openai",
    "baseUrl": "https://api.openai.com/v1",
    "apiMode": "openai"
  },
  {
    "id": "anthropic",
    "baseUrl": "https://api.anthropic.com",
    "apiMode": "anthropic",
    "headers": {
      "anthropic-version": "2023-06-01"
    }
  },
  {
    "id": "ollama",
    "baseUrl": "http://localhost:11434",
    "apiMode": "ollama"
  }
]

3. Define models

Each model must point at a provider:

"omnichat.models": [
  {
    "id": "gpt-5.4",
    "provider": "openai",
    "context_length": 128000,
    "max_completion_tokens": 8192,
    "vision": true
  },
  {
    "id": "gpt-5.4",
    "provider": "openai",
    "configId": "reasoning",
    "reasoning_effort": "high",
    "max_completion_tokens": 16384
  },
  {
    "id": "claude-sonnet-4",
    "provider": "anthropic",
    "max_tokens": 8192,
    "thinking": {
      "type": "enabled",
      "budget_tokens": 4096
    }
  },
  {
    "id": "llama3.1:70b",
    "provider": "ollama",
    "num_ctx": 32768,
    "temperature": 0.4
  }
]

4. Add or edit a provider

Run:

OmniChat: Edit Provider

This flow lets you:

Pick an existing provider and edit its apiMode, baseUrl, and API key
Add a new provider with the same form

Provider API keys are stored in VS Code Secret Storage under omnichat.apiKey.<providerId>.

5. Mount the provider into Copilot

Run:

Chat: Manage Language Models

Then:

Add a new language model group
Choose OmniChat
Enter the providerId you want to mount

That group will now expose only the models belonging to that provider.

Configuration Reference