On May 19, 2026, the official account of Google's developer-oriented terminal AI tool Gemini CLI announced on X platform that it will discontinue the Gemini CLI service for individual users, and all development resources will be unified into the Google Antigravity agent development platform. The platform has four independent entry points: Antigravity 2.0 Desktop, Antigravity CLI, Antigravity SDK, and Antigravity IDE.

The new Antigravity CLI has been rewritten in Go. It retains core features of the original Gemini CLI including Agent Skills, Hooks, sub-agents, and extensions (now renamed Antigravity plugins), while adding new capabilities for multi-agent orchestration and asynchronous workflows. According to Google, the rewritten CLI delivers faster response speeds, and complex multi-agent tasks can run in the background without blocking terminal sessions. It shares a unified agent scheduling backend with Antigravity 2.0 Desktop, and all future core feature updates will be rolled out synchronously across the entire platform.

The migration rules clearly distinguish between individual and enterprise users:

- Individual users with Google AI Pro/Ultra subscriptions, users of the free version of Gemini Code Assist for consumers, and users of Gemini Code Assist for GitHub must complete the migration within 30 days. Starting June 18, 2026, the old Gemini CLI will stop responding to requests, the GitHub version will no longer accept new installations, and will be gradually phased out. Google has already published [migration technical documentation](https://antigravity.google/docs/gcli-migration), and will release a step-by-step video tutorial later. Users can also download the latest version of Antigravity CLI directly from the [official channel](https://antigravity.google/download).

- Enterprise users are not affected by this adjustment. The original Gemini CLI service and API keys will continue to be supported, and enterprises can also choose to migrate to the Antigravity platform on a voluntary basis.

Google noted that Gemini CLI, launched in 2025, has accumulated millions of users in one year, received over 100,000 GitHub stars, and merged more than 6,000 community pull requests. The core reason for this adjustment is that developer workflows have shifted from single-agent to multi-agent collaboration, and the original architecture of Gemini CLI could no longer meet demand. A unified platform will speed up iteration and deliver a smoother experience for users.

Within hours of the announcement being published, a large number of user feedback appeared in the comment section:


Some users were unable to log in after updating to Antigravity CLI, getting a pop-up error about account settings, with the error indicating that the code assistance backend request failed.


Multiple other users have encountered the same login failure, and some also reported login request timeouts.


One user said they had been using Gemini CLI as a portable AI development environment on Termux, but the new Antigravity CLI does not support Termux, and they hope the official team can release a compatibility update as soon as possible. One developer mentioned that they had just finished creating a course on Gemini CLI ready for launch, and now all the content has to be scrapped.

Some netizens also complained about Google's product naming tradition: Google has launched so many independent products over the years including Bard, Gemini App, AI Studio, Vertex, Gemini CLI, Jules, and Antigravity, and now it says it will unify into one platform with four entries — that's so "Google-style". Some netizens joked that the name Antigravity CLI sounds like something out of a sci-fi movie, but they support the direction of this consolidation.

Many users have also put forward specific feature requests: support for Workspace account billing, unified usage limits across multiple products, clear rules for automated calls to avoid account bans, support for third-party OAuth integration, and a unified usage statistics dashboard. Some users who have already started using the new CLI said that when paired with Gemini 3.5 Flash, it is very fast and delivers a better experience than the old version.

Some developers in the discussion thread also recommended an open source companion tool called Engraph, which builds a persistent codebase context layer for AI coding agents, recording project architecture conventions, historical decisions, testing rules and other information to avoid having to re-explain the project background to the AI every session. The tool is released under the MIT open source license, compatible with mainstream AI coding tools including Claude Code and Cursor, can be installed directly via npm, and its core architecture is as follows:

```

┌─────────────────────────────────────────────────┐

│ Your AI Coding Agent │

└──────────────────────┬──────────────────────────┘

│ /context-search, /context-extract,

│ /context-add, /context-verify, /context-commit

│

┌────────▼─────────┐

│ Skills │ Read, write, and verify context

└────────┬─────────┘

│ invokes

┌────────▼─────────┐

│ Engraph CLI │ engraph graph / lookup / recall /

│ │ search / validate

└────────┬─────────┘

│ reads & writes

┌───────────────────┼────────────────────────────┐

│ │ │

┌──▼──────────────┐ ┌──▼──────────────────────┐ ┌───▼─────────────────┐

│ Codegraph │ │ Context Files │ │ Contextual Commits │

│ (.engraph/ │ │ (.engraph/context/) │ │ (in git history) │

│ codegraph/) │ │ conventions/ │ │ scoped action lines │

│ │ │ verifications/ │ │ via /context-commit │

│ Auto-generated │ │ │ │ │

│ structural map │ │ Version-controlled YAML │ │ Captured per commit │

│ Not committed │ │ Reviewed in PRs │ │ Recalled by module │

└─────────────────┘ └─────────────────────────┘ └─────────────────────┘

```

As of now, the official product team has responded in the comment section, stating that they will answer user questions one by one and iterate on features based on community feedback in the future.