Session 05 — Agent Mode in the IDE¶
Module: Copilot in Practice Difficulty: Intermediate Prerequisites: Sessions 01–03 Duration: 3 hours (1 hr trainer content + 2 hr lab)
Overview¶
Use agent mode for bounded, multi-step work in VS Code and JetBrains. It can plan changes, edit related files, run approved commands, respond to failures, and use configured MCP servers. This session focuses on defining a clear task boundary and reviewing the resulting work.
Learning Outcomes¶
- Explain the difference between assistant and agent workflows
- Use agent mode in VS Code for multi-step, multi-file tasks
- Use the agent loop: plan, act, observe, review, and revise
- Know when to use agent mode, Chat, or inline suggestions
- Use TODO code actions to bridge IDE and cloud agent workflows
Session Materials¶
| Resource | Location |
|---|---|
| Lab exercises | lab/ |
| Lab starter files | lab/starter/ |
| Lab solution files | lab/solution/ |
Key Topics¶
- From autocomplete to assistant to agent
- The agent loop: plan, act, observe, review, revise
- Tool use: terminal commands, file operations, web search
- Multi-file changes in agent mode
- Trust boundaries and when to intervene
- TODO code actions for cloud agent delegation
- Chat vs. Agent mode decision framework