Subagents in Gemini CLI

Subagents extend Gemini CLI by delegating portions of a broad or complex request to dedicated expert agents that run in isolated context windows. Each subagent maintains its own system instructions, tool set, and execution history, allowing the primary session to stay concise while the subagents perform research, code exploration, analysis, tests, or other high‑volume operations. The results are summarized and returned as a single response, reducing context rot and keeping interactions fast and cost‑effective.

The feature supports parallel execution, so multiple subagents can work simultaneously on different aspects of a task. Users can rely on built‑in specialists or define custom agents using simple Markdown files with YAML front‑matter, placing them in personal or project directories. This enables automation of specific workflows, enforcement of coding standards, or adoption of tailored personas within a team’s repository.

Because subagents operate independently yet report back to the main orchestrator, developers and power users can manage large, multi‑step processes without overwhelming the primary CLI context, making the tool suitable for complex productivity scenarios.