On this page
Gemini CLI Review: Open-Source Tool Gated by Corp Paywalls
The once open-source Gemini CLI, which amassed over 100,000 GitHub stars, is no longer accessible to free, Pro, or Ultra users as of June 18, 2026. Only enterprise license holders retain full access, while all other users are pushed to a closed-source replacement with a 98% smaller free tier. This shift serves as a case study in how open-source AI tools get captured for enterprise monetization.
Gemini CLI was supposed to be the open-source terminal agent that democratized AI-assisted coding. Instead, it became a case study in how open-source labor gets captured by enterprise monetization. As of June 18, 2026, the tool stopped serving requests for free, Google AI Pro, and Ultra users — only enterprise license holders retain access. Everyone else got pushed to a closed-source successor with a 98% smaller free tier. That’s not a product transition. That’s a gate.
What Gemini CLI Actually Was Before the Shutdown
Before the June 18 cutoff, Gemini CLI was an Apache 2.0 open-source project that accumulated over 100,000 GitHub stars and more than 6,000 merged pull requests from external contributors, per Google’s own announcement. It offered 1,000 requests per day on the free tier — generous enough for daily development work — and supported Agent Skills, Hooks, Subagents, and Extensions. The tool integrated with VS Code, JetBrains, and CI/CD pipelines through standard shell commands.
The architecture was TypeScript-based, community-governed, and genuinely useful for large codebase analysis. For teams running monorepos or complex refactors, the 1 million token context window was a real differentiator. If you’re still relying on it for large codebase work, you’ll need to understand exactly what works after the shutdown — we covered that in our guide to Gemini CLI for large codebases.
The June 18 Cutoff: What Broke and Who Got Hurt
On June 18, 2026, Gemini CLI and the Gemini Code Assist IDE extensions stopped serving requests for Google AI Pro, Ultra, and individual free-tier users, according to the official deprecation notice. The gemini command is broken for affected accounts, and the Login with Google option has been removed for IDE extensions and Gemini CLI.
The disruption wasn’t graceful. CI/CD pipelines, GitHub Actions, and scripts calling the gemini command broke immediately with no grace period for consumer-tier users, as multiple reports confirm. Individual users reported hard mid-session 403 errors on active paid Google AI Pro and Ultra accounts on June 18 — imagine losing access mid-refactor with no warning. Some developers on Reddit described the exact moment: a “You do not have a valid license of this product” error while the tool was actively reading files and writing code.
Enterprise customers had a different experience entirely. Gemini Code Assist Standard and Enterprise subscribers retain full, uninterrupted access to IDE extensions and Gemini CLI after June 18. The open-source repository continues to receive active development — v0.47.0 shipped after the cutoff, and v0.49.0 followed, per the release history. The tool isn’t dead. It’s just not for you anymore.
Antigravity CLI: The Closed-Source Replacement
Google’s replacement is Antigravity CLI, a closed-source, Go-based binary available since May 19, 2026, per the transition announcement. Google says it retains “the most critical features of Gemini CLI: Agent Skills, Hooks, Subagents, and Extensions (now Antigravity plugins).” But the company also admits there’s no 1:1 feature parity at launch.
The Antigravity CLI public GitHub repository contains only a changelog, readme, and GIF — no application source code, as 7minAI reports. That’s a stark contrast to the Apache 2.0 repository with 6,000 community contributions that preceded it. You can’t audit the code, you can’t fork it, and you can’t verify what the tool is doing with your data or API calls.
The free tier economics collapsed. Antigravity CLI’s free tier is approximately 20 requests per week (resetting weekly), compared to Gemini CLI’s 1,000 requests per day — representing roughly a 98% reduction in free-tier volume. That’s not a migration. That’s a quota gutting. Heavy users report exhausting the weekly cap in a handful of prompts.
The Pricing Landscape: What Enterprise Access Actually Costs
If you want to keep using Gemini CLI without disruption, you need an enterprise license. Gemini Code Assist Standard is priced at $22.80 per user per month and Enterprise at $54 per user per month, billed annually, per The Negotiation Experts. A 50-developer team using Gemini Code Assist Standard would cost $13,680 per year in subscriptions alone (50 × $22.80 × 12).
But that’s not the full picture. Compute usage is metered separately on top of seat fees. Gemini Enterprise Business is $21 per user per month, Standard is $30 per user per month, and Plus is $50 per user per month on annual commitment, with compute usage metered separately, per Cobry’s pricing analysis. For active development teams, expect to stack $15–$40 per power user per month in compute costs on top of the seat fee.
Here’s how the options compare:
| Option | Cost | Access Level | Source Code |
|---|---|---|---|
| Gemini CLI (free tier) | $0 | Dead after June 18 | Apache 2.0 |
| Antigravity CLI (free) | $0 | ~20 req/week | Closed |
| Gemini Code Assist Standard | $22.80/user/month | Full Gemini CLI access | Apache 2.0 (enterprise) |
| Gemini Code Assist Enterprise | $54/user/month | Full Gemini CLI + priority | Apache 2.0 (enterprise) |
| Gemini Enterprise Plus | $50/user/month | Workspace + CLI | Closed |
The pattern is clear: the open-source tool still exists, but only behind enterprise paywalls. If you’re evaluating migration options, our Gemini CLI pricing breakdown covers the full arc and hidden billing risks.
The Contradiction at the Heart of Google’s Transition
Google claims the transition was driven by user needs for modern multi-agent workflows. The official blog post states user workflows “have simply outgrown those early days of 2025” and require the multi-agent orchestration capabilities of Antigravity to “split up the work and solve complex problems.” That sounds reasonable until you notice the contradiction: the open-source Gemini CLI remains fully supported and updated for enterprise customers. The repository continues to receive active post-June 18 2026 releases with feature updates and bug fixes. Enterprise users retain full, uninterrupted access to the tool for all their workflows.
If Gemini CLI was truly obsolete for advanced use cases, why does Google keep developing it for paying customers? The answer is that the tool isn’t obsolete — it’s just not monetizable at scale for consumer tiers. The open-source version was never intended as a permanent community tool. It was a free R&D and marketing vehicle to attract enterprise customers, with thousands of community contributions serving as unpaid labor to build a product that would be locked behind enterprise paywalls.
What This Means for Your Migration Decision
If you’re a non-enterprise user right now, you have three realistic paths. First, migrate to Antigravity CLI and accept the 98% quota reduction — fine for occasional use, untenable for daily development. Second, pay for a Google AI Pro subscription and hope the Antigravity free tier improves (no guarantees). Third, evaluate alternatives like Claude Code or Codex CLI that don’t tie access to enterprise license tiers.
For enterprise customers, the calculus is different. You keep Gemini CLI access, but you’re now locked into Google’s ecosystem with metered compute on top of seat fees. The open-source license is intact, but the infrastructure that makes the tool useful — API access, model quotas, authentication — is fully controlled by Google. If they change pricing or deprecate features again, your only recourse is the contract you negotiated.
The broader pattern here isn’t unique to Google. It’s what happens when open-source projects become enterprise acquisition funnels. The community builds the product, the vendor captures the value, and the replacement is always more restrictive. If you’re building long-term tooling decisions around AI coding assistants, assume that any free tier is a temporary subsidy — not a permanent feature.
The open-source Gemini CLI was a good tool. It’s still a good tool, if you can pay for it. The question is whether you’re comfortable building workflows on infrastructure you don’t control, priced by a vendor that’s already demonstrated willingness to pull the rug. That’s not a technical decision. It’s a risk management one.