In short
Google moved free and Google AI Pro or Ultra users from Gemini CLI to Antigravity CLI on 18 June 2026. Gemini CLI remains Apache-2.0 licensed and is maintained for enterprise Code Assist users and API or Vertex keys. So most individuals are really choosing between Antigravity CLI and Claude Code. Antigravity has a $0 tier; Claude Code starts on the $20 Pro plan.
What changed
The Gemini CLI repository now spells out a significant change:
- Free individual access and Google AI Pro and Ultra plans moved to Antigravity CLI (
agy) on 18 June 2026. - Gemini CLI is maintained for enterprise Gemini Code Assist workflows on Google Cloud.
- It recommends Antigravity CLI for Gemini 4 Argon and later models, and says Gemini CLI is not tuned for Gemini 4+ and shouldn't be used for benchmarks.
- Those who can still use it include organisations with Code Assist Standard or Enterprise licences, Code Assist for GitHub on Google Cloud, and users with legacy paid API or Vertex keys.
So a comparison between Gemini CLI and Claude Code now mainly applies to enterprise teams already on Google Cloud. For everyone else, the live question is Antigravity CLI vs Claude Code. See also Antigravity vs Cursor.
Side by side
Checked on 4 November 2026.
| Gemini CLI | Antigravity | Claude Code | |
|---|---|---|---|
| Who it's for | Enterprise Code Assist; API or Vertex keys | Individuals, plus business through Google Cloud | Individuals and teams |
| Entry cost | Licence or usage-based billing | $0 individual plan, with basic weekly limits | Pro, $20/month ($17 annually) |
| Higher tiers | Via your organisation | Google AI Pro, Google AI Ultra | Max $100; Ultra $200 |
| Business | Code Assist | From $30/seat/month, or pay as you go | Team, from $20/seat annually |
| Licence | Apache-2.0 | See vendor terms | Proprietary |
| Project instructions | GEMINI.md | See vendor docs | CLAUDE.md, or AGENTS.md |
If you're on Gemini CLI at work
If your organisation has Code Assist licences, Gemini CLI still fits. It's open source, and it supports context files and MCP servers. Being able to authenticate through Vertex AI is useful if your data governance is already built around Google Cloud.
Check three things, though:
- Which models do you want? The project recommends Antigravity CLI for Gemini 4 Argon and later.
- What's your organisation's plan? Ask how long your team intends to stay on this route, so you don't standardise on a tool that's being narrowed.
- Could your scripts move? If you've automated workflows around it, keep them simple and portable.
Where Google's tools are strong

A free starting point. The Antigravity individual plan costs nothing and includes a choice of agent models, plus unlimited Tab completions and Command requests, with basic weekly rate limits. That makes it easy to try on a real project. Note that limits can tighten as demand changes.
Very long context. Gemini models have been known for large context windows, which helps when you need to reason across lots of code or documents at once. Check the current limits for your model and plan, as Gemini's pricing also varies by prompt size on some models.
Price per token. Some of Google's Flash and Flash-Lite models are very inexpensive through the API, although prices are scheduled to change on certain models. See our LLM API pricing comparison.
Google Cloud integration. Spend caps, admin controls and centralised policy are available for business plans.
Where Claude Code is strong
Agent maturity and customisation. It's documented around CLAUDE.md files, skills, hooks, subagents and MCP, along with scheduled tasks and CI integration.

Surfaces. You can use it in the terminal, VS Code, JetBrains, a desktop app, the web and on mobile, and move sessions between them.
Predictable subscription. You pay a flat fee with usage limits, which many developers prefer to metered billing.
Portability of instructions. It can read an AGENTS.md file in place of CLAUDE.md, which makes it easier to share conventions across tools.
The trade-off is that it has no free tier, and it's centred on Anthropic's models.
Don't rely on old benchmarks
Earlier comparisons often pitted older model generations against each other, or described Gemini CLI's free tier, which no longer applies. The repository itself warns against using Gemini CLI for benchmarking Gemini 4 models. If you read a comparison, check its date, then test for yourself. See the method in Claude Code vs Codex:
- Pick three real tasks from your repository.
- Start from the same commit, on separate branches.
- Give each tool the same instructions file and prompt.
- Compare how much of the output you'd keep, how many corrections it took, and how much of your allowance it used.
Protect yourself from tool churn
This whole situation shows why you shouldn't tie your workflow too closely to one vendor's tool. A product can change its audience in a single announcement. So:
- Keep project conventions in plain markdown files.
- Keep automation in ordinary scripts, instead of relying on tool-specific features where you can avoid it.
- Keep your code review and tests independent of whichever agent wrote the code.
- Review your choice every few months.
We wrote about what these instruction files should contain in how to write an AGENTS.md file and how to write a CLAUDE.md file.
How to choose
- Enterprise team on Google Cloud with Code Assist: Gemini CLI remains available, but ask about Antigravity's roadmap.
- Individual who wants a free start: Antigravity's free plan.
- Heavy agent use and wants customisation: Claude Code, or compare with Antigravity on Google AI Pro or Ultra.
- Needs an open-source CLI with model flexibility: consider OpenCode or the Codex CLI.
If your team needs help picking and standardising on an AI coding workflow, book a call.
Common questions
Is Gemini CLI still free?
Not for new individual use. The project says free individual access moved to Antigravity CLI on 18 June 2026, and its README lists no free tier or Google AI plans for Gemini CLI. Antigravity has a $0 individual plan with weekly limits.
What replaced Gemini CLI?
For free users and Google AI Pro or Ultra subscribers, Antigravity CLI, run with the agy command. Gemini CLI itself continues for enterprise Code Assist workflows and for users with legacy paid API or Vertex keys.
Can I still use Gemini CLI with an API key?
The project lists users with legacy paid API or Vertex keys as still able to use it. Check the repository for current details. It also recommends Antigravity CLI for Gemini 4 Argon and later models.
Is Antigravity better than Claude Code?
Neither is universally better. Antigravity has a free starting tier and Google integration, and Claude Code has a mature set of customisation features across many surfaces. Results shift with model releases, so test both on real work.