Cheat Sheet Quick reference
English

Language

Quick reference

Terminal AI Coding CLIs Comparison: Claude Code vs. Codex CLI vs. Antigravity CLI (agy)

Side-by-side reference comparing the three major terminal AI coding agents (Claude Code, OpenAI Codex CLI, Google Antigravity CLI agy): CLI flags, slash commands, shortcuts, memory/Skills/MCP config paths, and security sandbox models.

· Published

Terminal-native coding agents have converged around autonomous repository editing, test execution, MCP/Skills extensibility, and sandboxed permissions. Yet switching between Claude Code, OpenAI Codex CLI, and Google Antigravity CLI (agy) exposes subtle differences in flags, slash commands, config paths, and security policies.

This comparison cheat sheet provides a side-by-side equivalence map for developers and teams working across all three CLIs.

🔗 Individual CLI Cheat Sheets: Claude Code CLI Cheat Sheet · OpenAI Codex CLI Cheat Sheet · Antigravity CLI (agy) Cheat Sheet · AI Agent Harness Engineering Cheat Sheet

1. Architecture & Ecosystem Overview

Dimension Claude Code (claude) Codex CLI (codex) Antigravity CLI (agy)
Vendor Anthropic OpenAI Google (Antigravity 2.0 platform)
Runtime & Evolution Node.js / TypeScript ecosystem Rust-based local + cloud hybrid CLI Go rewrite; replaces retired consumer Gemini CLI
Primary Model Lineup Claude 5 series (Opus 5.5 / Fable 5.2) GPT-6 / Codex models Gemini 3.8 series & multi-model routing
Repo Memory File CLAUDE.md AGENTS.md GEMINI.md / AGENTS.md
Skills / Templates .claude/skills/<name>/SKILL.md ~/.codex/prompts/ (Markdown templates) .agents/skills/<name>/SKILL.md
Signature Features Lifecycle Hooks, one-command /pr codex cloud / codex apply sync Native /worktree, /goal, /grill-me, voice F5

2. CLI Startup & Automation Command Equivalence

Task / Intent Claude Code Codex CLI Antigravity CLI (agy)
Start interactive TUI claude codex agy
Start interactive with prompt claude "prompt" codex "prompt" agy -i "prompt"
Non-interactive / CI run claude -p "prompt" codex exec "prompt" (or codex e) agy -p "prompt"
Resume last / specific session claude --resume [id] codex resume [--last | <id>] agy -c / agy --conversation <id>
Select model claude --model <name> codex -m <name> agy --model <name>
Set reasoning depth Session config / API effort Reasoning Effort in /model agy --effort <low..max>
Manage MCP via CLI claude mcp [list | add | remove] codex mcp [list | add] TUI /mcp or edit mcp_config.json
Plugin / Extension tooling Plugin marketplace / local plugins — agy plugin [install | list | validate]
Upgrade / Diagnostics claude update / claude doctor Package manager / codex --strict-config agy self-update / agy plugin import gemini

3. Interactive Slash Commands & Shortcuts Equivalence

Action Claude Code Codex CLI Antigravity CLI (agy)
Initialize project memory /init ➔ CLAUDE.md /init ➔ AGENTS.md Create GEMINI.md / AGENTS.md
Command help /help / autocomplete menu ? / /help
Settings & permissions /config · /permissions /permissions · /status /config / /settings
Switch model / effort /model /model /model · /effort
Plan before editing Prompted planning / modes /plan /plan · --mode plan
Code review & diff /review · /diff /review · /diff (codex review) Workspace diff review + /worktree
Compact context /compact [focus] /compact /compact
Clear / Branch / Undo /clear /fork /clear · /fork · /rewind / /undo
Resume past session /resume codex resume from shell /resume
Inspect cost & tokens /cost · /context /status /config / session footer
Manage MCP servers /mcp /mcp /mcp
Inline file reference @<path> @<path> (fuzzy search) @<path> (path autocomplete)
Inline shell execution !<command> !<command> !<command>
External editor for prompts Shift + Enter multiline Ctrl + G ($EDITOR) Ctrl + G
Abort / Clean exit Esc abort · Ctrl + D exit Ctrl + C abort or exit Esc Esc clear · Ctrl + D Ctrl + D exit

4. Config Files, Memory & MCP Paths Comparison

Config Layer Claude Code Codex CLI Antigravity CLI (agy)
Global config ~/.claude/settings.json ~/.codex/config.toml ~/.gemini/antigravity-cli/settings.json
Project config .claude/settings.json (.local.json) .codex/config.toml .agents/ directory & workspace rules
Global memory ~/.claude/CLAUDE.md ~/.codex/AGENTS.md ~/.gemini/AGENTS.md / GEMINI.md
Project memory ./CLAUDE.md ./AGENTS.md ./AGENTS.md / ./GEMINI.md
Global MCP config ~/.claude.json ~/.codex/config.toml (MCP section) ~/.gemini/config/mcp_config.json
Project MCP config ./.mcp.json .codex/config.toml .agents/mcp_config.json
Remote MCP gotcha Standard HTTP / SSE TOML server table Must use serverUrl (url fails silently)
Custom skills / commands .claude/commands/ · .claude/skills/ ~/.codex/prompts/*.md .agents/skills/<name>/SKILL.md · .agents/rules/

💡 Coexistence Best Practice: Maintain shared repository build, test, and style guidelines in AGENTS.md (natively loaded by both Codex CLI and agy) and reference or mirror core constraints in CLAUDE.md to prevent rule drift across teams.

5. Security Models & Best-Fit Selection

Dimension Claude Code Codex CLI Antigravity CLI (agy)
Permission model Tool-level allow / ask / deny + Hooks Dual-layer: --sandbox + --ask-for-approval Modes (default / accept-edits / auto / plan) + Sandbox
Unattended / Auto flag --dangerously-skip-permissions (sandbox only) --sandbox workspace-write --ask-for-approval never (--full-auto deprecated) --mode auto or /goal
Sweet spot Deep multi-turn reasoning (Fable 5.2 / Opus 5.5), deterministic Hook enforcement, GitHub PR workflows Fast local iteration, Codex Cloud async offload, strict OS-level sandbox boundaries Multi-subagent concurrency, native Git worktree isolation, Antigravity IDE/2.0 parity & plugin bundles

Related topics

#claude-code#codex-cli#antigravity-cli#ai-coding#ai-agent#cli#cheat-sheet
  • Google Antigravity CLI (agy) Commands & Configuration Cheat Sheet

    A compact reference guide for Google's terminal coding agent Antigravity CLI (agy): covering startup flags and subcommands, TUI slash commands and shortcuts, .agents/ and mcp_config.json configuration, plus a Gemini CLI migration checklist.

  • OpenAI Codex CLI Commands & Configuration Cheat Sheet

    A compact reference guide for OpenAI's terminal coding agent Codex CLI: covering installation, auth, CLI subcommands and flags, sandbox and approval policies, TUI slash commands, keyboard shortcuts, config.toml, and AGENTS.md.

  • Claude Code CLI Commands & Slash Commands Cheat Sheet

    A compact reference guide for Anthropic's official terminal AI coding agent Claude Code: covering CLI startup flags, interactive slash commands (context management, permissions, code review, MCP extensions), keyboard shortcuts, and CLAUDE.md configuration patterns.

  • AI Agent Harness Engineering Architecture & Runtime Cheat Sheet

    A comprehensive reference guide to AI Agent Harness engineering and runtime scaffolds: comparing bare LLMs against harnessed agents, exploring the 6 core pillars of harness architecture, detailing a 10-step closed-loop bug-fixing workflow, and comparing top industrial harnesses (Claude Code, Devin, SWE-agent, OpenHands) with 5 golden engineering rules.