Antigravity CLI vs Claude Code vs Codex CLI: Honest 2026 Pick

TL;DR

Antigravity CLI, Claude Code, and Codex CLI are the three terminal coding agents worth comparing in 2026, and after running all three side by side for three months, my pick hasn’t moved. Claude Code is still the one I reach for when code quality matters. Codex CLI is the value play: cheaper on tokens, and the only one of the three that’s open source. Antigravity CLI has the best multi-agent orchestration story, but its rate limits still make it hard to recommend below the $200/mo Ultra tier. All of this traces back to Google killing Gemini CLI for individual users on June 18, 2026. The open-source tool that hit 100K GitHub stars stopped serving requests unless you hold an enterprise license. Antigravity CLI is its closed-source, Go-based replacement, and it ships with aggressive quotas that lock free-tier users out for up to seven days after 20-30 minutes of use. Google swapped in a credits-based top-up model in July, but the capacity meter still can’t be trusted.

Google’s Forced Migration

Google announced Antigravity CLI at I/O 2026 on May 19. The same day, the Gemini CLI deprecation notice went live. The timeline was tight, and it played out exactly as scheduled:

• May 19: Antigravity CLI launched, available to everyone
• June 18: Gemini CLI stopped serving requests for Google AI Pro, Ultra, and free-tier users
• June 18+: Gemini Code Assist for GitHub stopped accepting new installations

Enterprise customers with Gemini Code Assist Standard or Enterprise licenses keep their Gemini CLI access indefinitely. Everyone else had to migrate or switch tools before the cutoff.

The backlash was immediate, and for good reason. Gemini CLI was open source under Apache 2.0. Antigravity CLI isn’t. The GitHub repository contains a changelog, a readme, and a demo GIF, but no actual source code. The rate limits on the replacement are drastically more restrictive than what Gemini CLI users had. And the 30-day migration window was enough time to install a binary, but nowhere near enough to rebuild CI pipelines, port custom extensions, or audit a closed-source tool’s data handling.

I ran Gemini CLI daily for the tutorial I wrote in May. It worked well for what it was: a Node.js-based terminal agent with Gemini models, MCP support, and a sensible free tier. Antigravity CLI is supposed to be the upgrade. In practice, the experience is more complicated.

Antigravity CLI: What You Actually Get

Installation is straightforward:

# macOS / Linux
curl -fsSL https://antigravity.google/cli/install.sh | bash

# Verify
agy --version

The binary is agy, not antigravity. Authentication uses browser-based Google Sign-In. It opens your default browser, you log in, and the token stores in your system keyring. On SSH sessions, it prints a URL you complete in a local browser.

Antigravity CLI shares its backend with the Antigravity 2.0 desktop app. Conversations started in the CLI can be exported to the desktop GUI. Skills, hooks, MCP servers, and context files (GEMINI.md, AGENTS.md) all carry over from Gemini CLI, though the config paths change:

# MCP config (global)
~/.gemini/antigravity/mcp_config.json

# MCP config (workspace)
.agents/mcp_config.json

# Skills (global)
~/.gemini/antigravity/skills/

# Migrate existing Gemini CLI extensions
agy plugin import gemini

The Go rewrite does feel faster. Startup time dropped from ~1.2s (Gemini CLI on Node.js) to around 300ms. Memory overhead is lower. The TUI is keyboard-optimized and responsive. On pure ergonomics, it’s an improvement.

Where things fall apart is the rate limiting.

The Rate Limit Problem

Antigravity CLI uses a metric Google calls “compute effort” instead of transparent token counts. A simple question might consume 1 unit. A multi-file refactoring task might eat 50. The UI shows a capacity percentage, but it’s unreliable. One XDA reviewer reported their meter showed “100% capacity remaining” right before a 7-day lockout triggered.

The tiers:

One update since this comparison first ran: in July 2026, Google reworked Antigravity’s quota into a pay-as-you-go “AI credits” top-up model layered on top of the subscription tiers. The exact refresh windows in the table above have shifted as a result, so treat the specific numbers as the launch-era mechanics — but the core complaint, unpredictable multi-day lockouts that the capacity meter fails to warn you about, is still all over the forums. Adding credits softens the ceiling; it doesn’t make the meter honest.

For comparison, Claude Code’s Max 5x plan ($100/mo) gives you a consistent 5x the Pro window every 5 hours with no weekly ceiling. Codex CLI’s Pro plan ($200/mo) gives 20x Plus limits. Neither tool has users reporting multi-day lockouts on paid tiers.

I hit the free-tier wall on my second day with Antigravity CLI. A refactoring session across four files, maybe 25 minutes of active prompting, triggered a full weekly lockout. The Pro tier lasted longer but I still bumped into cooldowns during a two-hour coding session. The quota system feels tuned for casual one-off questions and breaks down fast during real development sessions.

Head-to-Head Comparison

Architecture, Models, and Benchmarks

The “Default model” row is the current shipping default (August 2026). The SWE-bench and Terminal-Bench numbers below are the figures each tool reported when I first tested them in May 2026, on that generation’s defaults (Opus 4.7, GPT-5.4, Gemini 3.1 Pro). Every vendor has shipped newer models since, and all three numbers have moved up — I’ve kept the original scores as a like-for-like snapshot of the tools rather than a live leaderboard.

The open-source gap is significant. Codex CLI’s full codebase lives on GitHub under Apache 2.0. Claude Code isn’t open source either, but Anthropic publishes transparent token pricing, so you can predict costs. Antigravity CLI is closed-source with opaque “compute effort” billing. You can’t inspect what the tool sends to Google’s servers, and you can’t predict what a task will cost before you run it.

On SWE-bench, the 72% vs 80-81% split came down to the models: Gemini 3.1 Pro trailed Opus 4.7 and GPT-5.4 on complex code generation. I broke down those model-level differences in GPT-5.4 vs Claude Opus 4.7 vs Gemini 3.1 Pro for coding, and ran a current-generation rematch in Qwen 3.7 Max vs GPT-5.5 vs Claude Opus 4.8 for coding once the newer models landed. If Google brings a current-generation Pro model to the CLI (the paid tier still defaults to the older Gemini 3.1 Pro, and free users only get 3.6 Flash), that gap could narrow.

Token efficiency is impossible to compare directly because Antigravity CLI doesn’t expose token counts. I compared Claude Code and Codex CLI’s real-world token usage before. Codex uses roughly 4x fewer tokens for the same task. A lot of that gap traces back to how an agent searches the repo before it acts, which is where coding agents can waste half their tokens. Antigravity CLI’s “compute effort” metric makes the same comparison impossible.

Pricing

At the $20/mo entry point, all three tools are nominally equivalent. In practice, Claude Code’s Pro plan gives a predictable 5-hour window that resets cleanly. Codex CLI’s Plus plan similarly refreshes without drama. Antigravity CLI’s Pro tier refreshes every 5 hours but sits under a weekly ceiling that’s caused lockouts for multiple users.

At the $100/mo tier, Claude Code’s Max 5x is the most predictable option: 5x the Pro quota per window, no weekly cap, no multi-day cooldowns. Codex CLI’s Pro tier at that price is a permanent 5x bump over Plus now that the launch 10x promo has ended. Antigravity CLI’s Ultra entry offers 5x Pro limits but users still report occasional throttling.

If you’re paying $200+, all three tools give you generous quotas. But at that price point, predictability and transparancy become the differentiators. Claude Code and Codex CLI both show you exactly what you’re consuming. Antigravity CLI does not.

Multi-Agent Capabilities

Antigravity CLI’s strongest differentiator is the Agent Manager. The Agent Manager lets you run multiple agents simultaneously in a dashboard-style view. You can kick off a refactoring agent, a test-writing agent, and a documentation agent in parallel. Progress is visible in a unified TUI. Conversations can be handed off between CLI and the Antigravity 2.0 desktop app.

Claude Code has subagents too. I covered the practical setup in a separate guide, and if you want the orchestration codified rather than driven by hand, I built four production Claude Code workflow scripts that fan out agents deterministically. They run in parallel via the Agent tool. But Claude Code’s subagent model is programmatic: you spawn agents with prompts, they return results, and you manage the orchestration in your conversation. There’s no dashboard.

Codex CLI runs parallel tasks in sandboxed environments. Each task gets its own filesystem context. It’s effective for batch operations but lacks the visual orchestration layer.

If your workflow involves multiple concurrent agents, say running a migration on three services simultaneously, Antigravity CLI’s Agent Manager is the most polished interface for that. If you’d rather script that orchestration than drive it from a dashboard, a dedicated framework like the ones I compared in LangGraph vs CrewAI vs OpenAI Agents SDK gives you more control. But most daily coding work is sequential: read, think, write, test. The multi-agent advantage is real but narrow.

What I’d Actually Pick

After running all three for three months, my recommendations split along three axes:

Pick Claude Code if you care about code quality above everything else. Opus 4.8 (the current Max default) produces the cleanest, most idiomatic code of the three, and its SWE-bench lead translates to noticeably fewer first-pass errors. Pricing is transparent and the Max plans give predictable capacity. The trade-off is higher token consumption. Claude Code thinks out loud and burns tokens doing it. If you’re on the API, Sonnet 5 at $2/$10 per million tokens handles 90% of daily work.

Pick Codex CLI if you prioritize token efficiency, open-source transparency, or autonomous batch work. (If async batch work is the main draw, also consider Google Jules — it queues tasks and opens PRs while you sleep, with up to 60 concurrent slots.) Codex uses 4x fewer tokens than Claude Code for the same tasks, and its 77.3% Terminal-Bench score shows it’s strongest at shell-oriented work. The Apache 2.0 license means you can inspect every line of code your agent runs. At $20/mo, ChatGPT Plus is the best value in this comparison; the $100/mo Pro tier is now a permanent 5x bump over Plus (the launch 10x promo has ended), and OpenAI has since added an $8/mo Go tier for lighter use.

Pick Antigravity CLI if you’re already deep in Google’s stack (GCP, Gemini API, Antigravity 2.0 desktop) and the multi-agent orchestration fits your workflow. The Go rewrite is fast and the Agent Manager UI is well-designed. But go in with your eyes open about the rate limits. Below the Ultra tier, the lockouts are unpredictable enough that I wouldn’t rely on it as my primary tool for deadline-sensitive work.

Skip Antigravity CLI if you’re a current Gemini CLI user who valued the open-source codebase, transparent usage metrics, and generous free tier. The replacement is a downgrade on all three of those dimensions. Google knows this. Enterprise customers get to keep Gemini CLI indefinitely. Individual developers don’t get that option.

Migration Guide for Gemini CLI Users

Gemini CLI stopped serving individual accounts on June 18, 2026. If you’re only migrating off it now, or setting up a fresh machine, here’s the practical checklist:

Staying on Google’s stack (Antigravity CLI):

# Install
curl -fsSL https://antigravity.google/cli/install.sh | bash

# Import your Gemini CLI extensions
agy plugin import gemini

# Move MCP configs
cp ~/.gemini/settings.json ~/.gemini/antigravity/mcp_config.json
# (manual edit needed — serverUrl replaces url for remote servers)

# Move skills
cp -r ~/.gemini/skills/ ~/.gemini/antigravity/skills/

# Context files (GEMINI.md) work unchanged

Switching to Claude Code:

# Install
npm install -g @anthropic-ai/claude-code

# Create context file
# Rename your GEMINI.md to CLAUDE.md (same concept, different name)
mv GEMINI.md CLAUDE.md

# MCP config goes in .claude/settings.json

Switching to Codex CLI:

# Install
npm install -g @openai/codex

# Context file is AGENTS.md (same name Antigravity uses)
# MCP config via codex config

All three tools support MCP servers, so any MCP integrations you’ve built will carry over with minimal config changes. The main migration cost is rewriting context files and adjusting to each tool’s quirks.

FAQ

Is Antigravity CLI open source?

No. The GitHub repository contains documentation and a demo GIF, but no source code. This is a significant departure from Gemini CLI, which was fully open source under Apache 2.0. Codex CLI and OpenCode are the remaining open-source options among the major terminal agents.

Does Gemini CLI still work after June 18, 2026?

Only for enterprise users with Gemini Code Assist Standard or Enterprise licenses, or those using paid Gemini Enterprise Agent Platform API keys. Individual Pro, Ultra, and free-tier users lost access on June 18, 2026, as scheduled.

Can I use Antigravity CLI for free?

Technically yes, but the free tier is limited to roughly 20 agent requests per day, and users report weekly lockouts lasting up to 7 days after moderate usage. For anything beyond occasional one-off questions, you’ll need a paid plan.

How do Antigravity CLI rate limits compare to Claude Code and Codex CLI?

Claude Code’s paid plans (Pro $20/mo, Max $100-200/mo) use transparent token-based quotas that refresh every 5 hours with no weekly cap. Codex CLI’s Plus ($20/mo) and Pro ($200/mo) similarly refresh without multi-day lockouts. Antigravity CLI’s paid plans refreshed every 5 hours under a weekly ceiling that caused 6-10 day cooldowns for Pro-tier users; Google switched Antigravity to a pay-as-you-go credits model in July 2026, but the multi-day-lockout reports predate that change and have continued past it.

Which terminal agent has the best code quality?

Claude Code, now defaulting to Opus 4.8, scored highest on SWE-bench Verified (~81% on the Opus 4.7 generation I tested in May 2026) and won 67% of blind code-quality evaluations. Codex CLI (GPT-5.4 then, GPT-5.6 now) was close at ~80% on SWE-bench but scored lower on blind quality tests. Antigravity CLI with Gemini 3.1 Pro sat at ~72%. Every vendor has shipped newer models since, but Claude Code has held the code-quality edge in my daily use.

Sources

• Transitioning Gemini CLI to Antigravity CLI — Google Developers Blog — official deprecation and migration announcement
• Choosing Antigravity or Gemini CLI — Google Cloud Blog — Google’s own feature comparison
• Antigravity CLI Deep Dive — Agentpedia — technical details on installation, config, and architecture
• Google Antigravity limitation — XDA Developers — hands-on review documenting rate limit issues
• Bye-bye, Gemini CLI — The Register — community backlash and open-source concerns
• Claude API Pricing — Anthropic — current Claude model pricing (Opus 4.8 / Sonnet 5)
• GPT-5.6 — OpenAI — GA July 9, 2026; current default model for Codex CLI
• Google AI subscriptions — The Keyword — Antigravity’s July 2026 pay-as-you-go AI-credits model

Bottom Line

Google had something good with Gemini CLI: an open-source terminal agent with reasonable limits and a growing community. Antigravity CLI is technically better (Go is faster than Node.js, the multi-agent orchestration is real) but worse in every dimension that built trust: it’s closed-source, the rate limits are punishing, and the migration timeline was aggressive.

If you’re a Gemini CLI holdout only now setting up a replacement, the honest recommendation is: try Antigravity CLI for a day and see if the rate limits work for your usage pattern. If they don’t, Claude Code at $100/mo (Max 5x) gives you the best code quality with predictable capacity. Codex CLI at $20/mo gives you the best token efficiency with full source-code transparency. Claude Code and Codex CLI both have long track records; Antigravity CLI has had three months in the wild now, and the rate-limit story still hasn’t settled. Google’s July switch to a credits model reshuffled the quotas without fixing the underlying unpredictability.

I’ve switched back to Claude Code as my daily driver. The rate limit story has to change before I’d trust Antigravity CLI on a deadline. Whichever agent you pick, the Agent Client Protocol lets you run it in JetBrains, Zed, or Neovim through a single config file.