Gemini CLI stops serving Pro, Ultra, and free users today

If you run Gemini CLI on a Google AI Pro, Ultra, or free plan, it stops serving requests today.

This isn't a deprecation notice with a runway. June 18, 2026 is the cut-off. The terminal you've been scripting against, and the Gemini Code Assist IDE extensions tied to the same plans, stop answering. Google's replacement, Antigravity CLI, is available to everyone as of today — but it's a fresh install and a real migration, not a silent upgrade.

Here's what actually changed and what to do about it.

Who's affected — and who isn't

The split is by license, not by how heavily you use the tool.

Stops working today:

• Gemini CLI for Google AI Pro, Ultra, and free (Gemini Code Assist for individuals)
• The Gemini Code Assist IDE extensions on those same plans
• Gemini Code Assist for GitHub — no new installs on GitHub organizations, and requests stop being served in the following weeks

Keeps working:

• Gemini CLI via a Standard or Enterprise Gemini Code Assist license
• Gemini CLI via paid Gemini and Gemini Enterprise Agent Platform API keys

So if your access runs through a company license or a billed API key, nothing breaks today. If it runs through an individual Pro/Ultra subscription or the free tier, it does.

What Antigravity CLI is

Antigravity CLI is Google's go-forward terminal agent. It's built in Go for faster execution and leans on asynchronous workflows — orchestrating multiple agents on a task in the background rather than one foreground loop.

The features most people built habits around carry over:

• Agent Skills
• Hooks
• Subagents
• Extensions — now called Antigravity plugins

The one thing to plan around: Google is explicit that there is no 1:1 feature parity out of the gate. Most of your setup ports, but expect gaps, especially in extensions/plugins and anything you wired together yourself. Treat the migration as a re-install where you verify each piece works, not a guaranteed lift-and-shift.

What to do

1. Check which tier you're actually on. This determines whether you have a problem today or not. Individual Pro/Ultra/free → migrate. Company license or paid API key → you can move on your own schedule.
2. Install Antigravity CLI and re-run your core workflow before you trust it. Confirm your Skills, hooks, and subagents behave the same. The parts most likely to differ are the plugins (formerly extensions) and any custom orchestration.
3. Audit anything automated. CI pipelines, cron jobs, or scripts that shell out to gemini will fail silently once requests stop being served. Find them before they find you in a red build.
4. If you only ever used it interactively, the move is small — install, sign in, carry on. The friction is concentrated in the automated and customized setups.

The broader point

The useful lesson here isn't about Gemini specifically. It's that a terminal agent from one of the largest companies in the world got pulled out from under individual users with a hard date — and the people most exposed are the ones who built the deepest automation on top of it.

That's the cost of wiring a single vendor's CLI into the load-bearing parts of your workflow. It's not an argument against using these tools; they're too useful to skip. It's an argument for keeping the seams loose: know what would break if any one agent disappeared, and keep the migration cost in your head before you need it. Today it's Gemini CLI. The pattern will repeat.

The take

If you're on Pro, Ultra, or free, your Gemini CLI stops today — move to Antigravity CLI and verify each piece, because the parity isn't 1:1. If you're on a company license or paid API key, you have time, but the same migration is coming. Either way, the thing worth internalizing is the platform risk, not the specific tool.

---

For how a multi-tool, multi-agent workflow actually fits together — and how to keep it resilient when one piece changes — /coding goes deeper.