News

Google Adds “Plan Mode” to Gemini CLI to Support Safer Code Planning

• By John K. Waters
• March 12, 2026

Google has introduced a new planning feature for its Gemini command line interface that allows developers to analyze and plan changes to software projects before modifying code.

The feature, called plan mode, was announced in a post on the Google Developers Blog on Wednesday. It places the Gemini CLI in a read-only environment designed to help developers study a codebase, map dependencies, and outline changes without altering files.

According to the company, the mode is intended to reduce the risk of accidental edits while developers evaluate complex changes.

“With plan mode, Gemini CLI focuses first on analyzing your request, planning complex changes, understanding the code base or dependencies, and all this happens in a read-only mode safe from accidental changes or executions,” the blog post said.

When plan mode is active, the tool can navigate code repositories, search for patterns, and read documentation, but it cannot modify project files. The agent is limited to read-only tools and can only create or modify its own internal plans.

Google said the feature is designed for tasks such as researching database migrations or outlining new features. Developers can refine a plan through conversation with the system before switching to an editing mode that allows code changes.

The update also introduces an “ask_user” tool that allows Gemini CLI to pause its research and request clarification from developers.

“This allows the agent to pause its research and ask you targeted questions to clarify your goals or gather missing information,” the company said.

Google said the tool enables the system to present options or request additional details instead of making assumptions about a developer’s intent.

Plan mode can also connect to external data sources through read-only Model Context Protocol tools. The feature allows developers to bring in context from services such as GitHub issues, database schemas, or documents while maintaining a read-only environment for the codebase.

Google said the feature is available by default to all Gemini CLI users. Developers can activate it by entering the command “/plan” in the interface or by selecting it through the tool’s approval mode settings.

The company said the system can use higher reasoning models, such as Gemini 3.1 Pro, during planning tasks to support architectural decisions and generate detailed plans.

The blog post also noted that developers can disable the feature through the settings menu if they prefer a workflow that begins directly with code editing.