Google has added subagents to Gemini CLI, enabling developers to delegate subtasks like code analysis, research, or testing to specialized AI agents running alongside the main session. The primary agent acts as an orchestrator, while subagents operate in isolated environments and return summarized results, reducing context overload. Subagents can run in parallel for faster execution, though this introduces risks like conflicting code changes and higher usage costs. Developers can define custom subagents using Markdown files with YAML configuration, specifying roles, tools, and behavioral guidelines. Built-in subagents are also available. Early user feedback points to ongoing usability and stability concerns with the tool.
Sort: