Skip to main content
GitHub Copilot integration allows you to use your Copilot subscription with Avante.nvim. This provider leverages your existing Copilot authentication from copilot.lua or copilot.vim.

​
Prerequisites

You must have either copilot.lua or copilot.vim installed and authenticated before using this provider.

​
Quick Start

1

Install Copilot plugin

Choose one:copilot.lua (recommended):
{
  "zbirenbaum/copilot.lua",
  config = function()
    require("copilot").setup()
  end,
}
copilot.vim:
Plug 'github/copilot.vim'
2

Authenticate with GitHub Copilot

Follow the authentication flow in your Copilot plugin (usually :Copilot setup or similar).
3

Configure Avante

{
  "yetone/avante.nvim",
  dependencies = {
    "zbirenbaum/copilot.lua",
  },
  opts = {
    provider = "copilot",
  },
}

​
Configuration

​
Basic Configuration

providers = {
  copilot = {
    endpoint = "https://api.githubcopilot.com",
    model = "gpt-4o-2024-11-20",
    timeout = 30000,
    context_window = 64000,
    extra_request_body = {
      max_tokens = 20480,
    },
  },
}

​
Available Models

Copilot provides access to various models. You can list available models: 
:lua print(vim.inspect(require('avante.providers').copilot:list_models()))
Common models: 
providers = {
  copilot = {
    model = "gpt-4o-2024-11-20",
  },
}

​
Authentication

​
OAuth Token Location

Copilot authentication is stored in:
  • copilot.lua: ~/.config/github-copilot/hosts.json
  • copilot.vim: ~/.config/github-copilot/apps.json

​
Token Refresh

Tokens are automatically refreshed:
  • Refreshed 2 minutes before expiration
  • Background refresh every 28 minutes
  • Stored in ~/.local/share/nvim/avante/github-copilot.json

​
Manual Refresh

If you need to re-authenticate:
  1. Re-authenticate with your Copilot plugin
  2. Restart Neovim
  3. Avante will automatically pick up the new token

​
Response API

Copilot supports OpenAI’s Response API for certain models: 
providers = {
  copilot = {
    -- Automatically enabled for gpt-5-codex models
    use_response_api = function(opts)
      local model = opts.model
      return model and model:match("gpt%-5%-codex") ~= nil
    end,
  },
}

​
Response API Features

  • Encrypted reasoning: Reasoning content is encrypted
  • Function calling: Enhanced tool use
  • Full history: Copilot doesn’t support previous_response_id, always sends full conversation
Unlike OpenAI’s Response API, Copilot requires sending the full conversation history with each request.

​
Model Listing

List available Copilot models: 
local models = require('avante.providers').copilot:list_models()
for _, model in ipairs(models) do
  print(string.format("%s (%s)", model.display_name, model.id))
end
Model information includes:
  • id: Model identifier
  • display_name: Human-readable name
  • name: Full model name with provider
  • tokenizer: Tokenizer used
  • max_input_tokens: Maximum input size
  • max_output_tokens: Maximum output size
  • version: Model version

​
Advanced Configuration

​
Custom Endpoint

providers = {
  copilot = {
    endpoint = "https://api.githubcopilot.com",
  },
}

​
Proxy Configuration

providers = {
  copilot = {
    proxy = "http://proxy.example.com:8080",
    allow_insecure = false,
  },
}

​
Request Headers

Copilot uses specific headers for authentication: 
-- Automatically set by Avante:
headers = {
  ["Authorization"] = "Bearer <token>",
  ["User-Agent"] = "GitHubCopilotChat/0.26.7",
  ["Editor-Version"] = "vscode/1.105.1",
  ["Editor-Plugin-Version"] = "copilot-chat/0.26.7",
  ["Copilot-Integration-Id"] = "vscode-chat",
  ["Openai-Intent"] = "conversation-edits",
}

​
Troubleshooting

If Copilot authentication fails:
  1. Verify Copilot plugin is installed: 
    :Copilot status
  2. Re-authenticate with Copilot: 
    :Copilot setup
  3. Check OAuth token exists: 
    cat ~/.config/github-copilot/hosts.json
Tokens are auto-refreshed, but if you see expiration errors:
  1. Restart Neovim to trigger token refresh
  2. Re-authenticate with Copilot if needed
  3. Check token file: ~/.local/share/nvim/avante/github-copilot.json
If a model isn’t available:
  1. List available models: 
    :lua print(vim.inspect(require('avante.providers').copilot:list_models()))
  2. Ensure you’re using a valid model ID
  3. Check your Copilot subscription includes the model
Error: “You must setup copilot with either copilot.lua or copilot.vim”
  1. Install copilot.lua or copilot.vim
  2. Authenticate with the plugin first
  3. Restart Neovim

​
Limitations

  • Cost: While included with Copilot subscription, be mindful of usage
  • Rate limits: Subject to Copilot’s rate limits
  • Auto-suggestions: Not recommended as auto_suggestions_provider due to high frequency (#1048)

​
Best Practices

Subscription

  • Requires active Copilot subscription
  • Individual or Business plans supported
  • Check limits in GitHub settings

Model Selection

  • Default models work well
  • Codex models for code-focused tasks
  • List models to see what’s available

Token Management

  • Tokens auto-refresh every 28 min
  • Check vim.g.avante_login for status
  • Re-auth if problems persist

Integration

  • Works alongside Copilot plugin
  • Shares authentication
  • Independent model selection

​
Example Configurations

{
  provider = "copilot",
  dependencies = {
    "zbirenbaum/copilot.lua",
  },
  providers = {
    copilot = {
      model = "gpt-4o-2024-11-20",
      timeout = 30000,
      extra_request_body = {
        max_tokens = 20480,
      },
    },
  },
}

Build docs developers (and LLMs) love

Get started for freeTalk to us