Skip to content

Visual Studio Code Agent Configuration Guide

Configure specialized AI agents in VS Code using GitHub Copilot

Overview

This guide provides practical examples and patterns for configuring GitHub Copilot agents in VS Code. For complete GitHub Copilot documentation, see the official GitHub Copilot docs.

What's here: Working configurations, integration patterns, and specialized agent examples you can copy and adapt.

Official Resources:

Documentation Structure

This guide is organized into focused topics:

🚀 Getting Started

  • Quick Start Guide - Get up and running in 5 minutes
  • Basic agent file setup
  • Your first agent invocation
  • Common agent examples

📋 Configuration Methods

  • Markdown-Based Agents - Declarative configuration (Recommended)
  • How markdown agents work
  • YAML frontmatter reference
  • VS Code workspace settings
  • Agent file organization

  • Complete agent examples

  • Programmatic SubAgents - The runSubagent tool

  • When to use runSubagent
  • How it works
  • Configuration strategies
  • Advanced examples

📚 Reference

  • Agent Examples Library - Ready-to-use agent configurations
  • Code Review Agent
  • Testing Specialist
  • Security Auditor
  • Documentation Generator
  • Performance Optimizer

  • And more...

  • Best Practices - Optimization and patterns

  • Agent design principles
  • Permission configuration
  • Temperature selection
  • Multi-agent workflows

  • Team collaboration

  • Troubleshooting - Common issues and solutions

  • Agent not recognized
  • Permission issues
  • Performance problems
  • Configuration errors

Key Concepts

Two Configuration Approaches

Create specialized agents as markdown files with YAML frontmatter:

---
name: reviewer
description: Code review specialist
model: claude-sonnet-4.5
temperature: 0.1
permissions:
  read: true
  write: false
  execute: false
---

# Code Review Agent
You are a specialized code review agent...

Invoke with: @reviewer Check this code for issues

Best for: - Reusable agent behaviors - Team-wide standardization - Quick access to specialized "modes" - Version-controlled configurations

Method 2: runSubagent Tool

Programmatically delegate complex tasks to autonomous sub-agents:

runSubagent({
  description: "Research authentication patterns",
  prompt: "Comprehensive task instructions..."
})

Best for: - One-off complex research - Autonomous multi-step tasks - Dynamic task generation - Deep codebase analysis

Quick Comparison

Feature Markdown Files runSubagent
Setup Create .md file once No setup needed
Invocation @agentname Full prompt each time
Reusability High Low
Team Sharing Excellent N/A
Complexity Simple Advanced
Best For Consistent roles Complex one-off tasks

Getting Started

  1. New to VS Code agents? Start with the Quick Start Guide
  2. Want to create custom agents? See Markdown-Based Agents
  3. Need complex automation? Check out Programmatic SubAgents
  4. Looking for examples? Browse the Agent Examples Library

What You'll Learn

Throughout this guide, you'll discover how to:

  • ✅ Create specialized agents for different workflows
  • ✅ Configure agent behaviors, models, and permissions
  • ✅ Invoke agents efficiently in your daily development
  • ✅ Build multi-agent workflows for complex tasks
  • ✅ Share agent configurations with your team
  • ✅ Troubleshoot common configuration issues

Additional Resources

Ready to get started? Jump to the Quick Start Guide →

Last Updated: November 25, 2025
Repository: ai-dev