Plan mode restricts Gemini CLI to working with read-only tools and prevents it from modifying any files except its own internal plans. Credit: Shutterstock / Jay Fog Google has added a plan mode to Gemini CLI, its open-source software engineering agent that runs on the command line. With plan mode, announced March 11, Gemini CLI focuses first on analyzing a request, planning complex changes, and understanding the codebase or dependencies in a read-only mode that is safe from accidental changes or executions. Plan mode will also ask questions to clarify goals before proposing a strategy for a user to review. Plan mode is now is enabled by default for all users, according to Google. Users can enter plan mode at any time by typing /plan in the input box, pressing Shift+Tab to cycle through approval modes, or simply asking the agent to “start a plan.” Plan mode restricts Gemini CLI to a subset of read-only tools such as read_file, grep_search, and glob to validate assumptions, Google said. When active, the agent can navigate a codebase, search for patterns, and read documentation, but it cannot modify any files except for its own internal plans. Users can ask Gemini CLI to “research how to migrate this database” or “plan a new feature,” and dependencies will be mapped out and a solution proposed without risk of eager code changes. A new ask_user tool has been introduced, allowing the agent to pause its research and ask targeted questions to clarify a user’s goals or gather missing information. Generative AIArtificial IntelligenceDevelopment ToolsSoftware Development