github github
stars 81
issues 3
subscribers 2
forks 10



22 hours ago

[!IMPORTANT] This plugin is provided as-is and is primarily developed for my own workflows. As such, I offer no guarantees of regular updates or support and I expect the plugin's API to change regularly. Bug fixes and feature enhancements will be implemented at my discretion, and only if they align with my personal use-cases. Feel free to fork the project and customize it to your needs, but please understand my involvement in further development will be intermittent. To be notified of breaking changes in the plugin, please subscribe to this issue.

:sparkles: Features

  • :speech_balloon: A Copilot Chat experience in Neovim
  • :electric_plug: Adapter support for many generative AI services
  • :robot: Agentic workflows to improve LLM output
  • :rocket: Inline code creation and modification
  • :sparkles: Built in actions for specific language prompts, LSP error fixes and code advice
  • :building_construction: Create your own custom actions for Neovim
  • :floppy_disk: Save and restore your chats
  • :muscle: Async execution for improved performance

:camera_flash: Screenshots

:zap: Requirements

  • The curl library installed
  • Neovim 0.9.2 or greater
  • (Optional) An API key for your chosen generative AI service

:package: Installation

Install the plugin with your package manager of choice:

-- Lazy.nvim
  dependencies = {
    "nvim-telescope/telescope.nvim", -- Optional
      "stevearc/dressing.nvim", -- Optional: Improves the default Neovim UI
      opts = {},
  config = true

-- Packer.nvim
  config = function()
  requires = {
    "nvim-telescope/telescope.nvim", -- Optional
    "stevearc/dressing.nvim" -- Optional: Improves the default Neovim UI

:wrench: Configuration

You only need to the call the setup function if you wish to change any of the defaults:

  adapters = {
    anthropic = "anthropic",
    ollama = "ollama",
    openai = "openai",
  strategies = {
    chat = "openai",
    inline = "openai",
  saved_chats = {
    save_dir = vim.fn.stdpath("data") .. "/codecompanion/saved_chats", -- Path to save chats to
  display = {
    action_palette = {
      width = 95,
      height = 10,
    chat = { -- Options for the chat strategy
      type = "float", -- float|buffer
      show_settings = true, -- Show the model settings in the chat buffer?
      show_token_count = true, -- Show the token count for the current chat in the buffer?
      buf_options = { -- Buffer options for the chat buffer
        buflisted = false,
      float_options = { -- Float window options if the type is "float"
        border = "single",
        buflisted = false,
        max_height = 0,
        max_width = 0,
        padding = 1,
      win_options = { -- Window options for the chat buffer
        cursorcolumn = false,
        cursorline = false,
        foldcolumn = "0",
        linebreak = true,
        list = false,
        signcolumn = "no",
        spell = false,
        wrap = true,
  keymaps = {
    ["<C-s>"] = "keymaps.save", -- Save the chat buffer and trigger the API
    ["<C-c>"] = "keymaps.close", -- Close the chat buffer
    ["q"] = "keymaps.cancel_request", -- Cancel the currently streaming request
    ["gc"] = "keymaps.clear", -- Clear the contents of the chat
    ["ga"] = "keymaps.codeblock", -- Insert a codeblock into the chat
    ["gs"] = "keymaps.save_chat", -- Save the current chat
    ["]"] = "keymaps.next", -- Move to the next header in the chat
    ["["] = "keymaps.previous", -- Move to the previous header in the chat
  log_level = "ERROR", -- TRACE|DEBUG|ERROR
  send_code = true, -- Send code context to the generative AI service? Disable to prevent leaking code outside of Neovim
  silence_notifications = false, -- Silence notifications for actions like saving saving chats?
  use_default_actions = true, -- Use the default actions in the action palette?


[!WARNING] Depending on your chosen adapter, you may need to set an API key.

The plugin uses adapters to bridge between generative AI services and the plugin (notably strategies). Currently the plugin supports:

  • Anthropic (anthropic) - Requires an API key
  • Ollama (ollama)
  • OpenAI (openai) - Requires an API key

Strategies are the different ways that a user can interact with the plugin. The chat strategy harnesses a buffer to allow direct conversation with the generative AI service. The inline strategy allows for output from the generative AI service to be written directly into a pre-existing Neovim buffer.

To specify a different adapter to the defaults, you can map an adapter to a strategy:

  strategies = {
    chat = "ollama",
    inline = "ollama"

[!TIP] To create your own adapter please refer to the ADAPTERS guide.

Configuring environment variables

You can customise an adapter's configuration as follows:

  adapters = {
    anthropic = require("codecompanion.adapters").use("anthropic", {
      env = {
        api_key = "ANTHROPIC_API_KEY_1"
  strategies = {
    chat = "anthropic",
    inline = "anthropic"

In the example above, we've changed the name of the default API key which the Anthropic adapter uses. Having API keys in plain text in your shell is not always safe. Thanks to this PR, you can run commands from within the configuration:

  adapters = {
    openai = require("codecompanion.adapters").use("openai", {
      env = {
        api_key = "cmd:op read op://personal/OpenAI/credential --no-newline",
    strategies = {
      chat = "openai",
      inline = "anthropic"

In this example, we're using the 1Password CLI to read an OpenAI credential.

Configuring adapter settings

Generative AI services have many settings such as model, temperature and max_tokens. In an adapter, these sit within a schema table and can be configured during setup:

  adapters = {
    anthropic = require("codecompanion.adapters").use("anthropic", {
      schema = {
        model = {
          default = "claude-3-sonnet-20240229",

[!TIP] Refer to your chosen adapter to see the settings available.

Edgy.nvim Configuration

The author recommends pairing with edgy.nvim for an experience similar to that of GitHub's Copilot Chat:

  event = "VeryLazy",
  init = function()
    vim.opt.laststatus = 3
    vim.opt.splitkeep = "screen"
  opts = {
    right = {
      { ft = "codecompanion", title = "Code Companion Chat", size = { width = 0.45 } },

Highlight Groups

The plugin sets the following highlight groups during setup:

  • CodeCompanionTokens - Virtual text showing the token count when in a chat buffer
  • CodeCompanionVirtualText - All other virtual text in the chat buffer

:rocket: Usage

The plugin has a number of commands:

  • :CodeCompanion - Inline code writing and refactoring
  • :CodeCompanionChat - To open up a new chat buffer
  • :CodeCompanionChat <adapter> - To open up a new chat buffer with a specific adapter
  • :CodeCompanionAdd - To add visually selected chat to the current chat buffer
  • :CodeCompanionToggle - Toggle a chat buffer
  • :CodeCompanionActions - To open up the action palette window

For an optimum workflow, the plugin author recommendeds the following:

vim.api.nvim_set_keymap("n", "<C-a>", "<cmd>CodeCompanionActions<cr>", { noremap = true, silent = true })
vim.api.nvim_set_keymap("v", "<C-a>", "<cmd>CodeCompanionActions<cr>", { noremap = true, silent = true })
vim.api.nvim_set_keymap("n", "<LocalLeader>a", "<cmd>CodeCompanionToggle<cr>", { noremap = true, silent = true })
vim.api.nvim_set_keymap("v", "<LocalLeader>a", "<cmd>CodeCompanionToggle<cr>", { noremap = true, silent = true })

-- Expand `cc` into CodeCompanion in the command line
vim.cmd([[cab cc CodeCompanion]])

[!NOTE] For some actions, visual mode allows your selection to be sent directly to the chat buffer or the API itself (in the case of inline code actions).

The Action Palette

[!NOTE] Please see the RECIPES guide in order to add your own actions to the palette.

The Action Palette, opened via :CodeCompanionActions, contains all of the actions and their associated strategies for the plugin. It's the fastest way to start leveraging CodeCompanion. Depending on whether you're in normal or visual mode will affect the options that are available to you in the palette.

[!TIP] If you wish to turn off the default actions, set use_default_actions = false in your config.

The Chat Buffer

The chat buffer is where you can converse with the generative AI service, directly from Neovim. It behaves as a regular markdown buffer with some clever additions. When the buffer is written (or "saved"), autocmds trigger the sending of its content to the generative AI service in the form of prompts. These prompts are segmented by H1 headers: user, system and assistant. When a response is received, it is then streamed back into the buffer. The result is that you experience the feel of conversing with your generative AI service from within Neovim.


When in the chat buffer, there are number of keymaps available to you:

  • <C-s> - Save the buffer and trigger a response from the generative AI service
  • <C-c> - Close the buffer
  • q - Cancel the stream from the API
  • gc - Clear the buffer's contents
  • ga - Add a codeblock
  • gs - Save the chat to disk
  • [ - Move to the next header
  • ] - Move to the previous header

Saved Chats

Chat buffers are not saved to disk by default, but can be by pressing gs in the buffer. Saved chats can then be restored via the Action Palette and the Load saved chats action.


If display.chat.show_settings is set to true, at the very top of the chat buffer will be the adapter's model parameters which can be changed to tweak the response. You can find more detail about them by moving the cursor over them.

Inline Code


You can use the plugin to create inline code directly into a Neovim buffer. This can be invoked by using the Action Palette (as above) or from the command line via :CodeCompanion. For example:

:CodeCompanion create a table of 5 fruits
:'<,'>CodeCompanion refactor the code to make it more concise

[!NOTE] The command can detect if you've made a visual selection and send any code as context to the API alongside the filetype of the buffer.

One of the challenges with inline editing is determining how the generative AI's response should be handled in the buffer. If you've prompted the API to "create a table of 5 fruits" then you may wish for the response to be placed after the cursor's current position in the buffer. However, if you asked the API to "refactor this function" then you'd expect the response to overwrite a visual selection. If this placement isn't specified then the plugin will use generative AI itself to determine if the response should follow any of the placements below:

  • after - after the visual selection
  • before - before the visual selection
  • cursor - one column after the cursor position
  • new - in a new buffer
  • replace - replacing the visual selection

As a final example, specifying a prompt like "create a test for this code in a new buffer" would result in a new Neovim buffer being created.

In-Built Actions

The plugin comes with a number of in-built actions which aim to improve your Neovim workflow. Actions make use of either a chat or an inline strategy. As mentioned above, the chat strategy opens up a chat buffer whilst an inline strategy will write output from the generative AI service into a Neovim buffer.

Chat and Chat as

[!TIP] Both of these actions allow for visually selected code to be sent to the chat buffer as code blocks.

Both of these actions utilise the chat strategy. The Chat action opens up a fresh chat buffer. The Chat as action allows for persona based context to be set in the chat buffer allowing for better and more detailed responses from the generative AI service.

Open chats

This action enables users to easily navigate between their open chat buffers. A chat buffer can be deleted (and removed from memory) by pressing <C-c>.

Agentic Workflows

[!WARNING] Agentic workflows may result in the significant consumption of tokens if you're using an external generative AI service.

As outlined by Andrew Ng, agentic workflows have the ability to dramatically improve the output of an LLM. Infact, it's possible for older models like GPT 3.5 to outperform newer models with traditional zero-shot inference. Andrew discussed how an agentic workflow can be utilised via multiple prompts that invoke the LLM to self reflect. Implementing Andrew's advice, the plugin supports this notion via the use of workflows. At various stages of a pre-defined workflow, the plugin will automatically prompt the LLM without any input or triggering required from the user.

Currently, the plugin only supports "reflection" (multiple prompts within the same application) and comes with the following workflows:

  • Adding a new feature
  • Refactoring code

Of course you can add new workflows by following the RECIPES guide.

Inline code

[!NOTE] The options available to the user in the Action Palette will depend on the Vim mode.

These actions utilize the inline strategy. They can be useful for writing inline code in a buffer or even refactoring a visual selection; all based on a user's prompt. The actions are designed to write code for the buffer filetype that it is initated in, or, if run from a terminal prompt, to write commands.

The strategy comes with a number of helpers which the user can type in the prompt, similar to GitHub Copilot Chat:

  • /doc to add a documentation comment
  • /optimize to analyze and improve the running time of the selected code
  • /tests to create unit tests for the selected code

Code advisor

[!NOTE] This option is only available in visual mode

As the name suggests, this action provides advice on a visual selection of code and utilises the chat strategy. The response from the API is streamed into a chat buffer which follows the display.chat settings in your configuration.

LSP assistant

[!NOTE] This option is only available in visual mode

Taken from the fantastic Wtf.nvim plugin, this action provides advice on how to correct any LSP diagnostics which are present on the visually selected lines. Again, the send_code = false value can be set in your config to prevent the code itself being sent to the generative AI service.

:rainbow: Helpers

Hooks / User events

The plugin fires the following events during its lifecycle:

  • CodeCompanionRequest - Fired during the API request. Outputs data.status with a value of started or finished
  • CodeCompanionChatSaved - Fired after a chat has been saved to disk
  • CodeCompanionChat - Fired at various points during the chat buffer. Comes with the following attributes:
    • data.action = close_buffer - For when a chat buffer has been permanently closed
    • data.action = hide_buffer - For when a chat buffer is hidden
    • data.action = show_buffer - For when a chat buffer is visible after being hidden
  • CodeCompanionInline - Fired during the inline API request alongside CodeCompanionRequest. Outputs data.status with a value of started or finished

Events can be hooked into as follows:

local group = vim.api.nvim_create_augroup("CodeCompanionHooks", {})

vim.api.nvim_create_autocmd({ "User" }, {
  pattern = "CodeCompanionInline",
  group = group,
  callback = function(request)
    print(request.data.status) -- outputs "started" or "finished"

[!TIP] A possible use case is for formatting the buffer after an inline code request


You can incorporate a visual indication to show when the plugin is communicating with a generative AI service in your Neovim configuration.


local M = require("lualine.component"):extend()

M.processing = false
M.spinner_index = 1

local spinner_symbols = {
local spinner_symbols_len = 10

-- Initializer
function M:init(options)
  M.super.init(self, options)

  local group = vim.api.nvim_create_augroup("CodeCompanionHooks", {})

  vim.api.nvim_create_autocmd({ "User" }, {
    pattern = "CodeCompanionRequest",
    group = group,
    callback = function(request)
      self.processing = (request.data.status == "started")

-- Function that runs every time statusline is updated
function M:update_status()
  if self.processing then
    self.spinner_index = (self.spinner_index % spinner_symbols_len) + 1
    return spinner_symbols[self.spinner_index]
    return nil

return M


local CodeCompanion = {
  static = {
    processing = false,
  update = {
    pattern = "CodeCompanionRequest",
    callback = function(self, args)
      self.processing = (args.data.status == "started")
    condition = function(self)
      return self.processing
    provider = " ",
    hl = { fg = "yellow" },

:gift: Contributing

I am open to contributions but they will be implemented at my discretion. Feel free to open up a discussion before embarking on a big PR.

:clap: Acknowledgements