GitHub Copilot CLI Agents for Spring Boot Developers

Copilot CLI does not run as one assistant. It delegates to specialized agents in separate context windows, the same idea behind Claude Code, so your main session stays focused instead of drowning in test output and file reads.

The full default roster:

• Explore — quick codebase analysis without polluting your main context

• Task — runs tests and builds, summary on success, full output on failure

• General purpose — complex multi-step work in a separate context

• Code review — surfaces only genuine issues, minimal noise

• Research — deep research across your code, related repos, and the web, with citations

• Rubber duck — a constructive critic, used automatically; you never call it

Research and Rubber duck are the two commonly left off the list. Research is the one that pulls from the web with sources, exactly what you want before a migration. Rubber duck you never invoke directly, but Copilot consults it on hard tasks on your behalf.

Why this matters if you write Spring Boot

The interesting question is not what the agents are, but which one handles which layer of a real request path, controller, service facade, repository, Kafka consumer, Spring Batch job, and what you actually type to route work to each. There is also a built-in /usage report that shows tokens, cost, and time, which matters because delegation multiplies token spend.

I worked through all of it, with a layer-by-layer prompt table and the cost reporting, in this week's deep dive on Product Cognizant:

[GitHub Copilot CLI's Default Agents for Spring Boot Developers →]

It pairs with the earlier piece on Claude Code's default agents, so if you switch between the two tools, you'll see exactly where they line up and where they differ.

Until next time, Kishore