Gemini CLI alternatives after the June 2026 change

We make Shuriken. We tried to be fair; corrections welcome at office@shuriken.sh.

Quick answer#

The right Gemini CLI alternative depends on how you signed in before 18 June 2026. With a personal Google account, free or on Google AI Pro or Ultra, Google's own replacement is Antigravity CLI. With a paid Gemini API key or a Code Assist licence, you can keep Gemini CLI. To leave Google's tools, look at Claude Code, Codex, Qwen Code or OpenCode.

That split comes from Google. Its 19 May 2026 announcement says Gemini CLI would stop serving Google AI Pro and Ultra and the free Gemini Code Assist for individuals on 18 June, while Code Assist Standard and Enterprise licences keep their access and "Gemini CLI will remain accessible via paid Gemini and Gemini Enterprise Agent Platform API keys" (Google Developers Blog). We read every page cited below on 24 October 2026.

Gemini CLI alternatives at a glance#

What replaced Gemini CLI for personal Google accounts#

Antigravity CLI did. The Gemini CLI repository opens with it: "Unpaid tier and Google One users: Gemini CLI was replaced by Antigravity CLI on June 18th, 2026" (GitHub). The command is agy, and on macOS or Linux it installs with curl -fsSL https://antigravity.google/cli/install.sh | bash (getting started).

It takes the account Gemini CLI now refuses. Google's install page has one line for it: "Complete the browser-based OAuth flow using your personal Google Account (Free, Pro, or Ultra plan)." The quota is shaped differently from Gemini CLI's old daily count, though. Google's plans page gives Google AI Ultra the largest quota, "refreshed every five hours", Pro a quota "refreshed every five hours until the weekly limit is reached", and accounts without Pro or Ultra "A meaningful quota, refreshed weekly". It gives no numbers.

Antigravity CLI is a different program, written in Go, not Gemini CLI under a new name. How the two compare on config, MCP servers, extensions and migration steps is in Gemini CLI vs Antigravity CLI, so we leave it out here.

Keeping Gemini CLI: a paid key or a Code Assist licence#

Gemini CLI still works if you pay Google by another route, and Google still lists two.

A Gemini Code Assist Standard or Enterprise licence comes through a company's Google Cloud account. Gemini CLI's quota page gives it 1,500 model requests per user per day on Standard and 2,000 on Enterprise.

A paid Gemini API key is the route for one person. Billing sits on the Google Cloud project behind the key: Tier 1 needs "an active billing account" linked to it, and Google applies rate limits "per project, not per API key" (Gemini API rate limits). The Gemini API has an unpaid tier, and Gemini CLI's quota page still lists 250 requests a day for an unpaid key on the Flash model only, but Google's announcement names only paid keys. We couldn't confirm that an unpaid key keeps working, so plan on a paid one.

What each sign-in allowed, before and after June, is in our Gemini CLI free tier guide.

Leaving Google: Claude Code, Codex, Qwen Code#

Claude Code#

Claude Code runs Anthropic's Claude models. You sign in with a Pro, Max, Team or Enterprise plan, or a Claude Console account billed per token (Anthropic). Claude's free plan doesn't include it (pricing). A company already on Google Cloud can also run it through Google Cloud's Agent Platform, set up with environment variables and no browser login. Our Claude Code vs Gemini CLI comparison puts the two side by side on limits and price.

Codex#

Codex comes with every ChatGPT plan, from Free to Enterprise, and OpenAI's pricing page describes the free one as "Explore Codex capabilities on quick coding tasks" (OpenAI). It also takes an OpenAI API key, billed at API prices. Like Gemini CLI, it's open source under Apache-2.0. If you already pay for ChatGPT, it's the quickest switch; Codex vs Gemini CLI has the limits and the config differences.

Qwen Code vs Gemini CLI#

Qwen Code is the closest thing to Gemini CLI you can install today, because it began as Gemini CLI's code. Its README says the project "was originally based on Google Gemini CLI v0.8.2", and that from Qwen Code v0.1 on "we stopped syncing with upstream and began independent development" (GitHub). The command is qwen, and npm install -g @qwen-code/qwen-code@latest installs it (Node.js 22 or later).

Its free sign-in is gone. Qwen Code's auth page says the "Qwen OAuth free tier was discontinued on 2026-04-15", so it now runs on an Alibaba Cloud Coding Plan (a monthly subscription), a Token Plan, a QwenCloud API key, or another provider's key. One of those providers is Gemini: its gemini provider reads GEMINI_API_KEY and GEMINI_MODEL (Qwen Code auth).

Keep your Gemini key, change the agent#

If what you liked about Gemini CLI was the Gemini models, a paid Gemini API key is the part that moves with you. Three agents besides Gemini CLI take one.

Antigravity CLI uses it once you set "modelProvider": "gemini" in ~/.gemini/antigravity-cli/settings.json and export GEMINI_API_KEY. Google warns that "Only setting a GEMINI_API_KEY environment variable on its own has no effect" (Antigravity).

Qwen Code takes it through its gemini provider, as above.

OpenCode takes it too. Its provider docs say it uses Models.dev to support "75+ LLM providers", and Models.dev's Google entry reads GEMINI_API_KEY, GOOGLE_API_KEY or GOOGLE_GENERATIVE_AI_API_KEY (Models.dev). OpenCode's own docs page documents Google Vertex AI, not the Gemini API key, so check the provider list in /connect before you count on it.

The key's limits don't change with the agent. They belong to the Google Cloud project, so two agents on one key share one quota.

Is there a free alternative to Gemini CLI?#

Three of these have a free way in on their vendor's own pages, as of 24 October 2026. Antigravity CLI signs in a free personal Google account, with a quota Google says is refreshed weekly. Codex is in ChatGPT Free for quick coding tasks. OpenCode's Console lists some models as free "for a limited time", though its setup steps still have you sign in, add billing details and credits, and copy an API key (OpenCode Console). Claude Code needs a paid plan or a Console account. Qwen Code's free tier ended in April 2026, so it runs on whichever provider's key you give it.

Where Shuriken is different#

Each agent above works from one sign-in or key per session, in its own terminal. Shuriken is a harness: it runs Gemini CLI, Antigravity, Claude Code, Codex, Qwen Code and OpenCode among its 67 coding agents, on your own computers, with your own plans and keys. Every chat but the Main chat gets its own branch in its own git worktree, so a Gemini CLI chat can work next to a Claude Code chat and a Codex chat in one project without editing the same files.

For Gemini CLI, each paid Gemini API key or licensed Google sign-in you add is its own source, and each sign-in keeps its own folder, so a second one never replaces the first. Shuriken's hot swapping moves your chats to another of your accounts or sources when one nears its limit. A Gemini CLI chat moves when the source it uses reports a limit: at the end of its current turn, it goes on with another of your Gemini sources (source switching). A chat that runs on one API key can move to your next key; a chat on a sign-in moves to a key only if you turn on "Switch to API keys", which is off by default and stops each fallback run at a $25 spend cap.

Antigravity chats use the sign-in agy already holds. Shuriken doesn't add them as sources, so it doesn't move them between accounts. Shuriken can't bring back Gemini CLI for a personal Google account either. Add only accounts and keys you own and are entitled to use; providers' terms may not allow sharing a subscription sign-in.

Pricing#

As published on 24 October 2026, in US dollars:

If you hold more than one paid Gemini key, or Gemini next to a Claude or ChatGPT plan, download Shuriken and add each as a source.