Tutorial: Set Up copilot-agent.nvim and Build a Flask To-Do App

This is the best place to start with copilot-agent.nvim.

In this tutorial, you will:

1. Install and configure the plugin
2. Open the chat UI inside Neovim
3. Use Copilot in agent mode to build a simple Flask to-do app
4. Ask Copilot to refine and test the app

---

Part 1 — Install and configure the plugin

Step 1 — Install the GitHub Copilot CLI

The plugin uses the official GitHub Copilot CLI runtime.

# macOS (Homebrew)
brew install copilot-cli@prerelease

# Or via npm
npm install -g @github/copilot@prerelease

Verify it:

copilot --version
# or
npx @github/copilot --version

Step 2 — Add the plugin with lazy.nvim

return {
  "ray-x/copilot-agent.nvim",
  build = ":CopilotAgentInstall",
  cmd = { "CopilotAgentChat", "CopilotAgentAsk" },
  keys = {
    { "<leader>aa", "<cmd>CopilotAgentChat<cr>", desc = "Copilot Agent Chat" },
  },
  config = function()
    require("copilot_agent").setup({
      service = {
        auto_start = true,
      },
      session = {
        enable_config_discovery = true,
      },
      chat = {
        diff_cmd = { "delta" },
      },
    })
  end,
}

Then run:

:Lazy sync
:CopilotAgentInstall
:checkhealth copilot_agent

Step 3 — Open the chat window

Inside Neovim:

:CopilotAgentChat

The prompt input opens automatically. You can start typing right away.

Step 4 — Switch to agent mode

Press i, or a to enter insert mode a prompt buffer will created under main chat output buffer.

Press <C-t> until the input statusline shows 🤖 agent.

In agent mode, Copilot can create files, run shell commands, and iterate on fixes.

Step 5 - Getting help

Press g? in the chat buffer to open the help menu with key bindings, slash commands, and session commands.

---

Part 2 — Build a simple Flask to-do app

Step 5 — Create a project directory

mkdir ~/flask-todo && cd ~/flask-todo
nvim .

Start from an empty directory.

Step 6 — Ask Copilot to scaffold the app

Start a new session with

:CopilotAgentChat

Send this prompt from prompt buffer:

Build a simple Flask to-do app in this directory.

Backend:
- Create a virtual environment and install flask
- Create app.py
- Keep data in memory for now
- Add these routes:
  - GET / serves the app
  - GET /api/tasks returns all tasks
  - POST /api/tasks creates a task
  - PUT /api/tasks/<id> updates a task
  - DELETE /api/tasks/<id> deletes a task
- Each task should have: id, title, completed
- Return JSON errors for invalid input and missing tasks

Frontend:
- Single-page app with plain HTML, CSS, and vanilla JavaScript
- Show the task list on the homepage
- Support add, toggle complete, edit, and delete
- Keep the styling simple and clean
- Make it responsive

Also create requirements.txt and a README.md with run instructions.

Approve the file writes and shell commands as Copilot works.

Step 7 — Run the app

Then ask:

Run the Flask app on port 5000

Open http://localhost:5000.

You should now have a working Flask to-do app.

Step 8 — Exercise the API

From a terminal:

curl http://localhost:5000/api/tasks

curl -X POST http://localhost:5000/api/tasks \
  -H 'Content-Type: application/json' \
  -d '{"title": "Buy milk"}'

curl -X PUT http://localhost:5000/api/tasks/1 \
  -H 'Content-Type: application/json' \
  -d '{"title": "Buy oat milk", "completed": true}'

curl -X DELETE http://localhost:5000/api/tasks/1

---

Part 3 — Refine the app

Step 9 — Ask Copilot for a small feature pass

Send:

Improve the to-do app:
- add a task counter
- add a "Clear completed" button
- improve empty-state messaging
- add smooth transitions when tasks change

Step 10 — Ask Copilot to add tests

Send:

Write pytest tests for the Flask to-do app.

Cover:
- GET / returns HTML
- GET /api/tasks returns JSON
- POST /api/tasks creates a task
- PUT /api/tasks/<id> updates a task
- DELETE /api/tasks/<id> deletes a task
- invalid task IDs return 404
- missing title returns 400

Add pytest to requirements.txt and run the tests.

Step 11 — Review the changes

Use:

:CopilotAgentDiff

---

What to do next

After this tutorial, continue with:

1. tutorial-weather-dashboard.md to learn how to use agents to refine an app
2. tutorial-custom-agent-weather.md to learn how to build with repo-local custom agents