On this page
Gemini CLI Pricing Explained: Free Tier to Forced Migration
Google shut down its free open-source Gemini CLI on June 18, 2026, cutting free tiers by 98% and funneling users to a closed-source, $100/month Antigravity successor. This breakdown covers the full pricing arc, hidden billing risks, and migration steps for teams and individual developers.
Google’s Gemini CLI pricing story isn’t a standard subscription tiers page — it’s a case study in how platform enclosure works when a company decides community goodwill has a revenue ceiling. Between mid-2025 and June 18, 2026, Google took an open-source terminal coding agent with 105,000 GitHub stars and 6,000 merged community contributions, offered it with a generous free tier, then sunset the entire open layer and funneled users into a closed-source, compute-restricted successor. If you’re trying to understand what happened, what it costs now, and whether the migration math works for your team, this breakdown covers the full arc.
What Gemini CLI’s Pricing Looked Like Before June 18
Gemini CLI’s original model was straightforward: authenticate with a personal Google account and you got 1,000 requests per day and 60 requests per minute at no cost. Paid tiers bumped that ceiling — Google AI Pro subscribers received 1,500 requests daily, while Google AI Ultra subscribers got 2,000. The tool itself was open-source under the Apache 2.0 license, written in TypeScript, and built through hundreds of independent contributor cycles. For a team evaluating terminal-based AI agents, the value proposition was hard to beat: a million-token context window, Google Search grounding, and a genuinely usable free tier that didn’t require a credit card.
That free tier mattered. It let developers run serious experiments — codebase exploration, debugging sessions, scripted automation — without committing budget. The quota documentation framed the tiers as a smooth ramp from experimentation to professional use. And for a while, that’s exactly how it worked.
Then Google I/O 2026 happened.
The June 18 Shutdown: What Stopped Working and Why
On June 18, 2026, Gemini CLI stopped serving requests for free, Google AI Pro, and Google AI Ultra subscribers. Enterprise Gemini Code Assist Standard and Enterprise license holders were explicitly exempted — they kept access. Everyone else got a hard cutoff. Any CI pipeline, cron job, shell script, or GitHub Actions workflow calling the gemini command simply broke. No grace period, no degraded mode. Just HTTP 410.
Google’s official framing was that developer workflows had “outgrown those early days of 2025” and now required “multiple agents communicating with each other.” The transition blog post described a unified backend as the technical justification. But the immediate practical effect was something else entirely: a free-tier collapse from 1,000 requests per day to roughly 20 per week in the replacement tool. That’s not a usage upgrade. That’s a 98% quota cut dressed up as a platform migration.
If you’re still running gemini in any automation, it’s broken. The question now is what to do about it.
Antigravity CLI: The Replacement Economics
Google launched a new $100/month AI Ultra plan at I/O 2026, specifically tailored for developers, technical leads, and knowledge workers. The company also reduced the top-tier AI Ultra plan from $250/month to $200/month. On paper, that looks like a price cut. In context, it functions as a migration toll.
Here’s the math that matters. The $100 plan offers 5X higher usage limits than the Pro plan. The $200 plan offers 20X higher limits. But these multipliers only matter if you can access the tool — and the free path to access no longer exists in any meaningful form. The Antigravity CLI free tier runs on a weekly compute-based cap rather than daily request limits, and community reports indicate the weekly quota exhausts in a handful of prompts. For any developer doing real work, the $100/month plan isn’t optional. It’s the entry price.
For teams, the costs scale fast. A 50-developer team migrating to Antigravity CLI on the $100/month AI Ultra plan would incur $60,000 annually in subscription costs — that’s 50 × $100 × 12, before any overage or compute charges. And unlike the original Gemini CLI, Antigravity CLI is closed-source. Its public GitHub repository contains only a changelog, a readme, and a GIF. No application code. The Apache 2.0 project that took 6,000 community pull requests now has no path for external contribution.
| Dimension | Gemini CLI (pre-June 18) | Antigravity CLI (post-June 18) |
|---|---|---|
| License | Apache 2.0 (open source) | Closed-source |
| Free tier | 1,000 requests/day | ~20 requests/week |
| Paid entry point | $0 (free tier usable) | $100/month |
| Binary | gemini (TypeScript) | agy (Go) |
| Config format | settings.json (inline MCP) | mcp_config.json (separate) |
| Skills directory | .gemini/skills/ | .agents/skills/ |
| CI/CD impact | — | Breaks gemini calls; manual migration required |
The Billing Transparency Problem Nobody’s Talking About
Beyond the quota cuts and the forced migration, there’s a separate issue that should concern any team running AI coding agents at scale. A GitHub issue documents a case where Gemini CLI told a user — directly and explicitly — that it used “Google’s internal quota” and was “not connected to your personal API key or your AI Studio projects.” That statement was false. The CLI silently charged the user’s GCP project at gemini-2.5-pro rates, and the user accumulated $4,000 in charges over two days with no billing alert until 48 hours later.
This isn’t a pricing explanation issue. It’s a trust issue. When an AI agent tells you it’s using a separate quota and then charges your production project instead, the problem isn’t the cost — it’s the opacity. Teams running Gemini CLI in automated workflows need to understand that the billing model was never as cleanly separated as the interface suggested. If you’re migrating to Antigravity CLI, audit your GCP billing alerts and set hard spend caps before running any automated agent workflows.
The Enclosure Cycle Pattern
What happened to Gemini CLI isn’t unique, but it’s a particularly clean example of a pattern I’ve started calling the Enclosure Cycle: a company uses open-source distribution and generous free tiers to build a community and validate a product category, then retires the open layer to force migration into a closed, enterprise-monetized platform. Google harvested community labor — 6,000 merged PRs, hundreds of contributors, a 105,000-star repository — to validate the terminal AI agent category. Then it sunset the open-source project to capture that value in a closed enterprise product, delivering a 98% free-tier reduction and silent CI/CD failures under the guise of a unified platform.
Google’s claim that users’ workflows “outgrown” Gemini CLI is a post-hoc rationalization. The company didn’t respond to demand for multi-agent workflows by adding multi-agent support to the open tool. It replaced the open tool with a closed one and eliminated the free tier that made the ecosystem viable. The community that built the product got a closed-source successor with no migration path that preserves what they built.
This matters beyond one tool’s pricing. If you’re evaluating AI coding agents as infrastructure — and you should be — the question isn’t just “what does it cost today?” It’s “what happens to my workflow if the vendor decides my free tier isn’t profitable anymore?” Gemini CLI’s pricing page looked stable for a year. Then it didn’t.
What Teams Should Do Now
If you’re running Gemini CLI in any capacity, your move depends on your situation:
- Free-tier individual users: You’re out of luck on the open-source path. Your options are the $100/month AI Ultra plan, a paid Gemini API key (which still works with the legacy CLI binary), or migrating to an alternative like Claude Code or Codex CLI. For a direct comparison of how Claude Code stacks up on pricing and capability, see our Claude Code vs Gemini CLI breakdown.
- Enterprise license holders: You’re exempt from the June 18 cutoff, but Google has made clear that Antigravity CLI is the forward path. Start planning your migration — the ecosystem support for Gemini CLI will decline.
- Teams with CI/CD automation: Audit every pipeline, script, and cron job that calls
gemini. The command is broken for affected accounts. Our migration guide coverage covers the config changes you can’t skip, including the silent MCP failure mode that breaks tool calls without obvious errors.
The Gemini CLI pricing story isn’t really about pricing. It’s about what happens when you treat a community-built open-source project as a customer acquisition channel for a closed enterprise product. The free tier was the hook. The enclosure was the business model. And the $100/month plan isn’t a discount — it’s the toll for crossing a bridge that used to be free.