Plan (/plan)

The /plan command facilitates structured planning, codebase exploration, and requirements discovery before jumping directly into code execution. For non-trivial tasks, /plan ensures complex changes are well-understood, properly scoped, and organized before files are modified.

---

Overview

Writing code before clarifying edge cases may lead to wasted effort, unnecessary diffs, or regressions. While standard prompts immediately begin modifying files, /plan introduces an intentional planning phase that explores the workspace, identifies dependencies, and surfaces architectural decisions for your review.

The /plan workflow helps bridge the gap between initial idea and code execution:

1. Exploration without side effects: The agent reads files, inspects call graphs, and analyzes dependencies without writing code prematurely.
2. Interactive discovery: If requirements are underspecified, the agent asks targeted questions to clarify trade-offs.
3. Reviewable artifact: The agent drafts a structured Implementation Plan artifact with clear task breakdowns and verification checkpoints.
4. User-guided execution: You retain full control to leave inline comments, request adjustments, or click Proceed to begin implementation.

---

How /plan works

When you invoke /plan, Antigravity runs a multi-phase planning and discovery workflow designed to produce an actionable implementation plan:

• Analysis and discovery: The agent analyzes your prompt, inspects workspace configuration, and examines affected files.
• Clarification and interviewing: If specifications are ambiguous or multiple technical approaches are viable, the agent initiates a focused interview using interactive question prompts and structured options.
• Structured plan creation: Once requirements are clear, the agent generates an Implementation Plan artifact.
• User review and approval: The agent presents the plan artifact in the review pane. You can review the proposed steps, leave inline annotations, or adjust individual tasks.

---

When to use /plan

The /plan command is especially valuable in the following development scenarios:

• Complex refactors: When restructuring modules, upgrading libraries, or refactoring interfaces across multiple files.
• Ambiguous requirements: When starting a feature with open questions, unknown constraints, or competing architectural approaches.
• High-risk changes: For database migrations, authentication overhauls, or security-sensitive patches where an upfront checklist is critical.
• Collaborative alignment: When you want to review and shape the agent’s strategy before any code modifications occur.

---

How to use /plan

You can invoke /plan with a specific task prompt or without arguments to start an open-ended discovery session.

Antigravity 2.0

Type /plan followed by your goal in the prompt input:

/plan Refactor our authentication middleware to support OIDC token verification with JWKS caching

To start an interactive discovery session for the current workspace context without a predefined prompt:

/plan

The agent will inspect your current active file and recent conversation history, then ask clarifying questions to help formulate a plan.

Antigravity CLI

Type /plan directly into the interactive prompt:

/plan Add rate limiting middleware to all public API endpoints using Redis token buckets

You can also run /plan without arguments:

/plan

The CLI enters planning mode, performs codebase analysis, and renders the structured plan directly in the terminal interface.

---

Interacting with plan artifacts

Plans generated by /plan use the standard Antigravity artifact format. You can interact with the plan in several ways:

• Inline comments: Highlight text or select specific plan steps to add targeted feedback.
• Scope adjustments: Ask the agent to break down tasks into smaller steps or remove optional work items before proceeding.
• Artifact review policies: Configure whether the agent pauses for plan approval using your project’s artifact review policy.
• Proceed: Click Proceed to transition the agent into execution mode. The agent follows the approved plan step-by-step, validating each milestone with tests.

For detailed instructions on artifact navigation and commenting workflows, see the Implementation plan guide.

---

Comparison of planning and reasoning modes

Antigravity provides complementary commands depending on task scope and complexity:

Command	Primary Focus	Best For	Plan Tier
/plan	Requirements discovery & reviewable plans	Scoping complex tasks, refactors, and multi-file changes before modifying code	All plans
/grill-me	Socratic design interviews	Deep requirement alignment and edge-case probing before drafting designs	All plans
/boost	Multi-agent deep reasoning	Hard concurrency bugs, race conditions, and algorithmic optimizations	Paid plans
/teamwork-preview	Collaborative agent teams	Long-horizon campaigns, repository migrations, and open-ended research	Paid plans

---

Next steps

Explore related planning and automation documentation:

• Implementation plan artifact: Learn how to review, comment on, and approve plan artifacts.
• Artifact review policies: Configure review behavior and auto-proceed settings.
• Slash commands catalog: Explore the complete catalog of slash commands.
• Boost deep reasoning (/boost): Leverage multi-agent reasoning loops for tough engineering challenges.
• Teamwork agent teams (/teamwork-preview): Run collaborative agent teams for repository-scale projects.