Gemini CLI Is Shutting Down: 3 Migration Paths You Should Consider
What happened to Gemini CLI?
Google announced at I/O on May 19, 2026: Gemini CLI will stop responding to individual users on June 18, 2026. Free accounts, Pro subscribers, Ultra subscribers — all cut off.
The official replacement is Antigravity CLI (agy), a closed-source Go binary. The Gemini CLI repo keeps its Apache 2.0 license, but the product is dead for personal use.
Enterprise customers (Google Cloud Code Assist Standard/Enterprise) and API key users are not affected.
Why is the community angry?
The GitHub Discussion #27274 has 247 downvotes against 4 upvotes. The frustrations are specific and well-documented.
Closed source betrayal. Gemini CLI had 100,000+ GitHub stars, 6,000 merged PRs, and hundreds of contributors. All that community work now flows into a proprietary product. Andrea Alberti, a 27-commit contributor, asked directly whether developers had “essentially been working for free on a codebase that will only be used in enterprises.” Another wrote bluntly: “We’re making the project closed source. That’s it.” The Linux Foundation spotlighted this pattern at Open Source Summit North America 2026 — Gemini CLI became exhibit A for their new Model Openness Tool.
Token consumption spike. One developer posted GCP billing screenshots comparing the same task on both tools. Flash 3.5 in Antigravity consumed dramatically more tokens than the same model in Gemini CLI. As they put it: “The token usage of Flash 3.5 is not only unreal, but also makes Antigravity 2.0 unusable, no matter how good or modern the tool is.”
Quota collapse. Gemini CLI’s old free tier allowed up to 1,000 requests per day. Antigravity’s free tier resets every 166 hours (~7 days). Reddit users report hitting the weekly cap within 4-5 chat turns. One wrote: “even with pro I got the usage limit in just 6 to 7 prompts, this is insane, earlier I used to make whole projects with Gemini CLI with only 13% quota reached.”
Missing features. No diff view for modifications. No .md project memory. Can’t paste screenshots from clipboard. Features that existed in Gemini CLI are simply gone.
MCP incompatibility. MCP servers configured for Gemini CLI don’t transfer cleanly. The UI detects them, but the agent ignores them.
API key login removed. Independent developers who authenticated via AI Studio keys can no longer log in without a Google account OAuth flow.
No documentation or bug tracker. There’s no discussion board, no issue tracker, no public roadmap for agy-cli. As one Hacker News commenter noted: “If the agent won’t tell you what its programming is, it’s not your agent.”
30-day deadline. As Sean Kim wrote: “Replacing a free, unlimited, open-source tool with a metered product on a 30-day deadline, before the replacement is ready. That is not a deprecation; it is a forced conversion event.”
What are your options?
You have three paths. Each has real tradeoffs.
Path 1: Switch to Antigravity CLI (Google’s official answer)
This is the path of least resistance if you’re committed to Google’s ecosystem.
# Download the binary
curl -fsSL https://antigravity.google/download | sh
# Authenticate
agy auth login
# Import old config
agy migrate --from-geminiYour skills directory moves from .gemini/skills/ to .agy/skills/. The gemini command becomes agy.
| Pros | Cons |
|---|---|
| Same models (Gemini Flash, Pro) | Closed source |
| One-command migration | MCP servers may break |
| Faster execution (Go binary) | Missing diff view |
| Official Google support | Low free quota (7-day refresh) |
| No community forum | |
| Higher token consumption |
Pick this if: You need Gemini models specifically, or your org is already on Google Cloud.
Path 2: Keep Gemini CLI with a paid API key
This is the sleeper option nobody talks about. The shutdown only affects individual user authentication. API key access via Google AI Studio continues working.
# Generate an API key at https://aistudio.google.com/apikey
export GEMINI_API_KEY="your-key-here"
# Gemini CLI respects this env var
gemini "refactor auth module"You keep the exact same workflow. Same binary, same config, same MCP servers. You just pay per token instead of using the free/Pro/Ultra quota.
| Pros | Cons |
|---|---|
| Zero migration effort | Pay-per-token cost |
| Existing MCP servers work | No more free tier |
| Same familiar interface | Gemini CLI won’t get new features |
| Open source (Apache 2.0) | You’re on an abandoned tool long-term |
Pick this if: You have working workflows you don’t want to break, and you’re fine paying for API access while you evaluate alternatives.
Path 3: Switch to a different tool entirely
The Gemini CLI shutdown is a natural forcing function to reconsider your tooling. The alternatives are mature:
| Tool | Language | License | Key Strength |
|---|---|---|---|
| Claude Code | Node.js | Proprietary | Best code understanding (200K context) |
| Codex CLI | Rust | Open source | Full sandboxing, OpenAI models |
| OpenCode | Go | Open source | Multi-provider, TUI interface |
| Kiro CLI | Node.js | Proprietary | AWS integration, skill system |
# Claude Code
npm install -g @anthropic-ai/claude-code
# Codex CLI
npm install -g @openai/codex
# OpenCode
go install github.com/opencode-ai/opencode@latest| Pros | Cons |
|---|---|
| Better tooling (in some cases) | Migration cost |
| Active development | Different model ecosystems |
| Community support | Learning curve |
| Not dependent on Google’s decisions | May need new MCP configs |
Pick this if: You’ve been unhappy with Gemini’s direction, or you want a tool with an active open community.
How to decide?
Ask yourself two questions:
- Do you need Gemini models specifically? If yes, Path 1 or 2. If no, Path 3.
- Do you have MCP servers and custom workflows? If yes, Path 2 buys you time. If no, migration cost is low either way.
The worst choice is doing nothing. June 18 is six days away.