GitHub Copilot Agents

​

What Are Copilot Agents?

​

Accessing and Invoking Agents

@workspace

​

Common Agents and Their Domains

​

Agents in Action: Mention-Based Workflow

1. Open Copilot Chat (Ctrl+Shift+P → Copilot: Open Chat)
2. Type @workspace and ask a question:
   @workspace Explain how this traffic simulation application works.
   
   The agent will scan your entire project and highlight relevant functions. For instance:
   def draw_traffic_light(surface, direction, state):
       """
       Draws a circular traffic light for a given direction at a fixed offset
       from the intersection.
       """
       positions = {
           'north': (CENTER_X, STOP_LINE['north'] - LIGHT_OFFSET),
           'south': (CENTER_X, STOP_LINE['south'] + LIGHT_OFFSET),
           'east':  (STOP_LINE['east'] - LIGHT_OFFSET, CENTER_Y),
           'west':  (STOP_LINE['west'] + LIGHT_OFFSET, CENTER_Y)
       }
       pos = positions.get(direction)
       color = LIGHT_COLORS[state]
       pygame.draw.circle(surface, color, pos, LIGHT_RADIUS)
   
3. Switch to the Terminal agent to install and run the app:
   pip3 install pygame
   python3 -m venv testenv
   source testenv/bin/activate
   pip install pygame
   python main2.py
   
   If you encounter an indentation error:
   (venv) jeremy@MACSTUDIO code % python main2.py
   Traceback (most recent call last):
     File "main2.py", line 220, in <module>
       main()
     File "main2.py", line 216, in main
       run_simulation()
   IndentationError: unindent does not match any outer indentation level
   
   Correct the indentation and rerun to launch the Pygame traffic light simulator.
4. For editor-specific queries, use the VS Code agent:
   @vscode How do I enable Copilot Chat experimental features?
   
   It will direct you to Settings > Editor > Copilot.
5. Generate tests across your codebase with Workspace:
   @workspace fix tests
   
   This produces comprehensive unit tests rather than file-scoped ones.

​

Experimental Agent Mode

1. Click the Agent Mode toggle in the lower-right corner of VS Code.
2. Request a new project scaffold:
   Create a Python API application.
   
   The agent will generate a virtual environment, install Flask, and produce app.py:
   from flask import Flask
   
   app = Flask(__name__)
   
   @app.route('/')
   def hello_world():
       return 'Hello, World!'
   
   if __name__ == '__main__':
       app.run(debug=True)
   
   And run:
   python3 -m venv venv
   source venv/bin/activate && pip install Flask
   source venv/bin/activate && python app.py
   
   * Running on http://127.0.0.1:5000/ (Press CTRL+C to quit)
   * Restarting with stat
   * Debugger is active!
   * Debugger PIN: 371-990-415
   127.0.0.1 - - [06/Mar/2025 14:45:21] "GET /" 200 -
   

Add an in-memory database with fields ID, name, and email.

​

Summary

• Mention-based Copilot Agents (@workspace, @terminal, @vscode, etc.) scan specific contexts to deliver targeted assistance.
• Experimental Agent Mode enables Copilot to autonomously generate code, run commands, and modify files.

​

Links and References

• GitHub Copilot Documentation
• VS Code Extension: GitHub Copilot