On this page
Gemini CLI for Large Codebases: What Works After Jun 18 2026
Google shut down free and paid consumer access to Gemini CLI on June 18, 2026 with no warning, breaking workflows for developers using the tool for large codebase analysis and monorepo refactors. Only enterprise license holders retain full access, while non-enterprise users must migrate to the closed-source Antigravity CLI or alternatives like Claude Code.
Google shut down free and paid consumer access to Gemini CLI on June 18, 2026 — no grace period, no warning for mid-session users. If you’re running a large codebase and relying on the gemini command for codebase exploration, cross-package refactors, or monorepo-wide analysis, your workflows just broke. The tool’s 1M-token context window and directory-level context injection made it genuinely useful for large-project work. Now the question is what still functions, what doesn’t, and what your actual options are.
For a full breakdown of the pricing arc and hidden billing risks, see Gemini CLI Pricing Explained: Free Tier to Forced Migration. If you need a step-by-step migration walkthrough, Gemini CLI Best Practices: What to Do Before June 18, 2026 covers the essentials.
The Shutdown Reality: Who Keeps Access
Gemini CLI stopped serving requests for Google AI Pro, Ultra, and free-tier users on June 18, 2026, with no grace period, per Simon Carter’s coverage. CI/CD pipelines, GitHub Actions workflows, cron jobs, and any automation calling the gemini command on those tiers stopped working that day.
The exemption is narrow: Gemini Code Assist Standard and Enterprise license holders retain full access to Gemini CLI, according to the official GitHub discussion. If your organization has enterprise seats, nothing changes for you. If you’re an individual developer, a small team on Pro/Ultra, or running automation under a personal account — you’re in the affected bucket.
The replacement is Antigravity CLI, a closed-source, Go-based rewrite announced May 19, 2026. Google acknowledged that Antigravity CLI does not have 1:1 feature parity with Gemini CLI at launch, as reported by 9to5Google. For teams running large-codebase workflows, this gap matters more than it does for single-file edits.
What Gemini CLI Could Do for Large Codebases
Before the shutdown, Gemini CLI had accumulated a set of capabilities specifically designed for working across large projects. These are the features you’re losing access to — or at least, losing the free/affordable path to.
Context window and file size. Gemini CLI supports a 1,048,576-token context window (1M tokens), per TechJackSolutions’ guide. It can process inline files up to 100MB, after a PR updated the hardcoded limit from 20MB to match the Gemini API’s new threshold. For monorepos with large generated files or bundled assets, that ceiling matters.
Context management features. Gemini CLI v0.35 introduced Just-In-Time (JIT) context loading, and v0.38 added Context Compression, both designed to improve performance with large codebases. These aren’t marketing labels — they directly affect how much of your project the agent can hold in a single session before it starts dropping or summarizing context.
Plan mode and worktree isolation. Plan mode, a read-only analysis mode, became enabled by default in Gemini CLI v0.34 (March 17, 2026), per Verdent Guides. Git Worktree support for isolated agent workspaces landed in v0.36 (April 1, 2026). Together, these let you spin up an agent that analyzes your entire repo without touching your main branch — useful for large-scale refactors where you want the agent to reason about the codebase before proposing changes.
Subagents for parallel work. The subagents feature was added to Gemini CLI on April 20, 2026, enabling parallel task handling by distributing work across specialized agents, as announced on the Google Developers Blog. For a large codebase, this means you could delegate frontend updates, test generation, and documentation to separate agents simultaneously rather than burning through a single context window sequentially.
Cross-package reasoning in monorepos. Official documentation includes a large codebase analysis example demonstrating cross-package reasoning across a monorepo using directory-level context injection (@packages/...), per PR #24104. This is the workflow that teams working in monorepos actually depend on — mapping a PolicyEngine implementation across core, cli, and a2a server packages in a single session.
The Migration Path and Its Traps
Google’s official migration guide states the interactive migration takes about 15 minutes, and the agy plugin import gemini command automates most config and plugin transfers. The steps are: install the agy binary, run agy plugin import gemini, rename GEMINI.md to AGENTS.md, move .gemini/skills/ to .agents/skills/, and update scripts to use agy.
But user reports tell a different story for automation-heavy workflows. Some MCP configuration fields fail silently during migration from Gemini CLI to Antigravity CLI, without throwing errors, per the Digital Applied migration guide. If your CI pipeline depends on MCP servers for database schemas or internal API docs, a silent config failure won’t show up until a pipeline breaks in production.
The quota model also changed. Before shutdown, Gemini CLI’s free tier for individual Google account users provided 1,000 requests per day, per Gemini CLI’s documentation. The new weekly compute-based quota is a different model entirely, and heavy users are already reporting it as a significant downgrade. For teams running automated codebase analysis or CI-integrated review, a weekly cap with no daily granularity means burst workflows will hit walls unpredictably.
The Open Source Value Transfer Problem
Here’s the structural issue that makes this more than a routine product sunset. Google accepted 6,000+ community contributions to the Apache 2.0 Gemini CLI repository over nearly a year, accumulating 100,000+ GitHub stars. Then on June 18, 2026, the company cut off all non-enterprise users and funneled them toward a closed-source replacement that, by its own admission, lacks feature parity.
The repository remains publicly available under Apache 2.0, but it’s functionally non-functional for non-enterprise users after June 18, 2026, due to backend access restrictions. You can read the code. You can fork it. You can’t use it for anything useful without enterprise-tier API access.
This is what I call the open-source value transfer pattern: a company uses an open-source license to attract community labor, builds a product with that labor, then captures the value exclusively for paying enterprise customers. The license is intact. The community is locked out. The tool they helped build no longer serves them.
What Your Options Actually Are
If you’re working with large codebases and just lost Gemini CLI access, here’s the landscape:
| Tool | Free Tier | Context Window | Open Source | Large Codebase Support |
|---|---|---|---|---|
| Gemini CLI (Antigravity CLI) | No (free tier ended Jun 18) | 1M tokens | No (closed-source Go binary) | Partial — lacks feature parity at launch |
| Claude Code | No | 1M tokens | No | Strong — subagents, mature context management |
| Codex CLI | No | Varies by model | — | Early-stage |
If you have enterprise seats, nothing changes. Keep using Gemini CLI. Evaluate Antigravity CLI in parallel. You’re fine.
If you’re on a paid Gemini API key, you retain access. The quota pool is shared with Antigravity, so monitor your usage carefully.
If you need open-source or self-hosted, the landscape is thin. The original Gemini CLI code is Apache 2.0 but non-functional without Google’s backend. You’re looking at building your own harness around a Gemini API key, which is viable but means maintaining your own context management and tooling layer.
For a deeper comparison of Claude Code and Gemini CLI — including agentic workflow capabilities and vendor trust records — check out Claude Code vs Gemini CLI: The 2026 Reality Check.
The Bottom Line
Gemini CLI was one of the few free tools that could hold a large codebase in context and reason across packages. That capability didn’t disappear — it just got locked behind enterprise licensing and a closed-source replacement that isn’t feature-equivalent yet. If your workflow depended on cross-package analysis, monorepo context injection, or automated CI integration, you’re now facing a migration with no clean drop-in solution.
The pattern here is worth watching. Google isn’t the first company to use open-source licensing to capture community value for enterprise products. But Gemini CLI is a particularly clean case study: 6,000 contributions, 100,000 stars, then a hard cutoff with no grace period. If you’re evaluating any open-source AI coding tool today, ask yourself what happens to your access if the company decides your tier isn’t profitable enough. The Apache 2.0 license protects the code. It doesn’t protect your workflow.