Cooking with Agents in VS Code — Liam Hampton, Microsoft

2026-05-21 · Source: AI Engineer · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Intermediate, long

Summary

Liam Hampton from Microsoft presents VS Code as a unified entry point for managing diverse AI agents, including GitHub Copilot agents. The presentation categorizes agents into three types: local agents for hands-on tasks like writing tests, background agents (e.g., GitHub Copilot CLI with Git worktrees) for 50/50 human-in-the-loop tasks like UI creation, and cloud agents for hands-off tasks such as documentation generation. A demonstration showcased a simultaneous multi-agent workflow within VS Code, where a local agent wrote unit tests using "Claude Opus 46", a background agent developed a front-end UI, and a cloud agent generated open-source repository files. The talk highlights VS Code's customization modal for configuring agents, skills, instructions, and hooks, supporting both first-party and third-party services like Claude. Cloud agents operate securely in GitHub Actions, leveraging MCP servers for extended context and built-in safeguards.

Key takeaway

For AI Engineers and Software Engineers seeking to optimize development workflows, integrating diverse AI agents within VS Code offers a powerful solution. You should utilize VS Code's unified interface to orchestrate local, background, and cloud agents based on task complexity and desired human involvement. This approach reduces cognitive load and accelerates development cycles, allowing you to simultaneously manage tasks like UI generation, documentation, and unit testing. Explore the customization options to tailor agents to your specific project needs.

Key insights

VS Code serves as a single, customizable hub for orchestrating diverse AI agents, optimizing developer workflows.

Principles

• Agent types should align with task involvement.
• Simultaneous agent use boosts productivity.
• Agent customization enhances utility and control.

Method

The article describes a multi-agent workflow: use a background agent for UI creation, a cloud agent for documentation, and a local agent for unit tests, all managed within VS Code's unified interface.

In practice

• Configure VS Code's agent modal for custom agents/skills.
• Use Git worktrees with background agents for isolated development.
• Explore aka.ms/awesomecopilot for agent customization examples.

Topics

• AI Agents
• VS Code Integration
• GitHub Copilot
• Multi-Agent Orchestration
• Mono Context Protocol
• Developer Workflow

Best for: AI Engineer, Software Engineer

Related on AIssential

• Claude Code Agent View IS INSANE! Huge New Update Introduces /goal, sessions, & More! — Cloud Code's new features centralize agent management and enable autonomous, goal-driven execution for developers.
• AI agents aren't replacing software engineering but expanding it far beyond code, researchers argue — AI agents expand software engineering beyond code, necessitating new methods for "semi-executable artifacts."
• Use subagents and custom agents in Codex — OpenAI Codex now offers subagents and custom agents for modular, specialized AI-driven development tasks.
• The Accidental Orchestrator — Effective AI-driven development requires a structured approach and human orchestration of AI agents, not just tool use.
• multica-ai / multica — Multica integrates coding agents as autonomous teammates, managing their lifecycle from task assignment to skill reuse.
• A Harness for Every Task: Putting a Team of Claudes on One Job — Dynamic Workflows overcome single-agent limitations by orchestrating fresh-context sub-agents via a code-based plan.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by AI Engineer.