Gemini CLI is not a sleepy terminal wrapper around a model. ToolVitals sees 29 release events in 30 days, 30 GitHub releases in 90 days, 105,258 GitHub stars, a 100 shipping score, and a 98 ToolVitals score. The interesting part is not just that Google is shipping it hard. The interesting part is that Gemini CLI is being maintained at high velocity while Google is also steering consumer users toward Antigravity CLI.
What Gemini CLI is built to be
The official Gemini CLI site describes the product in plain terms: build, debug, and deploy with AI from the terminal. The page says developers can query and edit large codebases, generate apps from images or PDFs, and automate complex workflows with Gemini 3. The install path is also direct: npm install -g @google/gemini-cli.
The GitHub repository uses a similar framing. It calls Gemini CLI an open-source AI agent that brings Gemini into the terminal, with lightweight access to Gemini and a short path from prompt to model. That matters because this is not positioned as a hosted IDE, a notebook, or a cloud-only code assistant. The core interface is the command line.
The repository excerpt lists the features Google wants developers to notice: a free tier of 60 requests per minute and 1,000 requests per day with a personal Google account, Gemini 3 models with a 1M token context window, built-in tools for Google Search grounding, file operations, shell commands, and web fetching, MCP support for custom integrations, and terminal-first workflows.
Gemini CLI is also not just an interactive prompt. The first-party repository page says it supports non-interactive scripts, workflow automation, conversation checkpointing, custom GEMINI.md context files, and GitHub workflow integration. The GitHub Action section names pull request reviews, issue triage, on-demand assistance by mentioning @gemini-cli, and custom scheduled or on-demand workflows.
That product shape is clear. Gemini CLI is a terminal AI agent with codebase access, model access, shell access, web access, and integration hooks. In practical terms, it competes less with a chat UI and more with the growing class of agentic developer tools that sit inside the development loop.
The openness classification is also clear. ToolVitals marks Gemini CLI as OSI-approved OSS with an Apache-2.0 license signal. That supports calling it open source. The repository excerpt backs that up by describing it as Apache 2.0 licensed.
The surprise is the transition, not the velocity
The official Gemini CLI homepage now carries a blunt notice: unpaid tier and Google One users will be moved to Antigravity CLI on June 18. The page links to Google’s Developers Blog post, “An important update: Transitioning Gemini CLI to Antigravity CLI.”
That post frames the change as product consolidation. Google says it is unifying agentic developer tooling around Google Antigravity, described in the post as its agent-first development platform. Gemini CLI is transitioning to Antigravity CLI for consumers because Google wants one product for what it calls multi-agent reality.
The blog post gives useful context for the scale of Gemini CLI before the transition. It says Gemini CLI reached millions of users, 100,000+ GitHub stars, 6,000 merged pull requests, and hundreds of contributors. ToolVitals does not use those blog numbers as scoring inputs here, but they line up with the GitHub repository scale ToolVitals measured: 105,258 stars and very high shipping activity.
Google also says Antigravity CLI is available to everyone as of May 19, 2026. The post says Antigravity CLI is built in Go, supports asynchronous workflows, and can orchestrate multiple agents in the background. It also says Antigravity CLI shares the same agent harness as Antigravity 2.0.
This changes the buyer question. If you are evaluating Gemini CLI as a stable long-term interface for consumer-tier use, the answer is no longer just “look at the stars and releases.” You need to ask which access path your team uses. The blog post says enterprise users on Gemini Code Assist Standard, Gemini Code Assist Enterprise, Gemini Code Assist for GitHub through Google Cloud, paid Gemini API keys, or Gemini Enterprise Agent Platform API keys are treated differently from consumer users.
That is the center of the story. Gemini CLI looks very healthy in repository terms, but some users are being told to migrate. High activity and migration pressure can both be true.
What the ToolVitals numbers say
ToolVitals gives Gemini CLI a hot score of 229.2, a health score of 95, a shipping score of 100, and an overall ToolVitals score of 98. Data confidence is 99. Those are not soft vibes. They are strong signals that the project is active, visible, and heavily maintained.
The star count is the loudest adoption signal: 105,258 GitHub stars. Stars are imperfect. They can reflect launch hype, brand gravity, curiosity, or bookmarks rather than production usage. Still, in developer tools, a six-figure star count means the project has crossed from niche utility into mainstream developer awareness.
The release cadence is even more operationally important. ToolVitals records 29 release events in 30 days and 30 GitHub releases in 90 days. That is not occasional maintenance. That is a release machine.
The repository’s own release policy explains part of that cadence. The local first-party excerpt says new preview releases are scheduled weekly on Tuesdays at UTC 23:59, stable releases are scheduled weekly on Tuesdays at UTC 20:00, and nightly releases are published daily at UTC 00:00 from main branch state. That means the release count is partly structural, not purely a measure of feature breakthroughs.
That distinction matters. A daily nightly channel can inflate raw release volume compared with a project that ships only stable tags. ToolVitals avoids pretending that every release event is equal to a major product change. The right interpretation is narrower: Gemini CLI has an active release pipeline, automated release discipline, and frequent public artifacts.
The recent release notes support that reading. The v0.48.0 nightly release on June 13 includes atomic update work in MCP tool discovery, Vertex AI model mapping fixes, Antigravity CLI migration documentation and commands, empty resume session cleanup, dependency automation, tool output formatting, workflow logging changes, zero-quota fail-fast behavior, and shell wrapper parsing fixes.
That is not one theme. It is maintenance across core agent behavior, model routing, migration support, CI, tool formatting, and terminal edge cases. Messy? A bit. Healthy? Yes.
Release notes show a project hardening the agent surface
The v0.46.0 release on June 10 is a good example of the kind of work Gemini CLI is getting. The release notes list a PTY resize hardening fix against native crashes, a fix preventing a spam loop when preferredEditor is invalid, a transition to a Flash GA model when an experiment flag is present, CI workflow changes, and patch cherry-picks from preview releases.
Those are not flashy features. They are the stuff that makes a terminal agent tolerable in real use. PTY resizing, editor preference handling, model selection, and CI policy are all boring until they break. Then they become the whole product.
The v0.47.0 preview release continues the same pattern. It includes backend definition handling for 3.5 Flash, auto mode changes for 3.5 Flash when the relevant flag is enabled, policy fixes for EBUSY fallback and TOML parse recovery, atomic MCP tool discovery updates, Antigravity transition banner work, Antigravity migration commands, and browser agent documentation cleanup.
MCP shows up repeatedly. The repository excerpt positions MCP as the route for custom integrations, and the v0.48.0 nightly includes atomic update work in MCP tool discovery. That is a specific signal about where terminal agents are headed. The terminal is not enough by itself. The agent needs controlled access to tools, servers, context, and project-specific capabilities.
The Antigravity references show another direction. Gemini CLI releases now include migration commands and transition banner tuning. That is product plumbing for a user migration. Maintainers are not just shipping agent improvements. They are also managing the handoff from one Google terminal experience to another.
For maintainers, this is the key risk to manage publicly: users can confuse high release volume with roadmap clarity. The release notes are doing useful work, but the project would benefit from unusually explicit channel guidance. Stable, preview, nightly, enterprise-supported Gemini CLI, consumer Antigravity CLI, paid API-key usage, and free-tier migration are a lot for one README to carry.
Where Gemini CLI sits against nearby tools
ToolVitals related-tool data puts Gemini CLI in a tight cluster of highly active developer tools. React Email has a 227.2 hot score, 19,323 stars, a 94 shipping score, and 29 release events in 30 days. Mem0 has a 226.7 hot score, 58,511 stars, a 100 shipping score, and 21 release events in 30 days.
Gemini CLI leads both on stars, with 105,258, and edges them on hot score at 229.2. React Email matches Gemini CLI’s 29 release events in 30 days, which is a nice reminder that release velocity alone does not explain the whole ranking. Gemini CLI’s combination of release activity and unusually large GitHub audience is the sharper signal.
Electric SQL is the outlier in raw shipping motion. ToolVitals records 108 release events in 30 days and a 100 shipping score for Electric SQL, with 10,230 stars and a 225.1 hot score. That comparison is useful because it separates two kinds of heat. Electric SQL is hyperactive in release count. Gemini CLI is active while also carrying a much larger public developer audience.
Composio and Dyad sit nearby too. Composio has 28,760 stars, a 94 shipping score, and 7 release events in 30 days. Dyad has 20,619 stars, a 98 shipping score, and 8 release events in 30 days. Against that group, Gemini CLI looks less like an experimental side project and more like a flagship developer tool under active operational pressure.
What the data does not prove
ToolVitals can see public signals: stars, releases, release events, score inputs, SSL and uptime when tracked, and repository activity signals. It cannot see whether Gemini CLI produces good code for your codebase. It cannot see user satisfaction, retention, revenue, enterprise renewal rates, internal Google priorities, or support burden.
It also cannot turn the Antigravity transition into a simple health penalty. The project can be healthy and still be changing shape. In fact, the inspected first-party sources show both sides: Gemini CLI is active, documented, Apache-2.0 open source, and shipping fixes, while Google is also directing consumer users toward Antigravity CLI.
Web search was unavailable during this research run because the configured search backend returned a credit block. That limits cheap third-party corroboration. The evidence here comes from ToolVitals metrics, local first-party excerpts, direct inspection of the official Gemini CLI site, the GitHub repository, GitHub release pages, and the Google Developers Blog transition post.
A skeptical engineering lead should treat Gemini CLI as a strong candidate only after checking access mode. If your team depends on free individual access, the June 18 transition notice is not a footnote. If your team uses enterprise Gemini Code Assist or paid API-key access, the official blog says access remains different, but you should still test the migration path and document which CLI your automation expects.
Recommendation
If your team wants a terminal-first AI coding agent with real open-source licensing, strong public adoption, MCP integration, GitHub workflow hooks, and fast maintenance, evaluate Gemini CLI. The ToolVitals numbers justify the evaluation: 105,258 stars, 29 release events in 30 days, a 100 shipping score, and a 98 overall score.
Do not evaluate it as if nothing is changing. For consumer-tier usage, start with Antigravity CLI and compare behavior against Gemini CLI before wiring it into scripts. For enterprise or paid API-key usage, Gemini CLI still deserves a serious trial, especially if your team lives in terminals and needs agent workflows that can touch files, shell commands, web context, and MCP tools.
The practical path is simple. Install it in a non-critical repo, define a GEMINI.md, test codebase querying, MCP discovery, shell operations, checkpointing, and non-interactive scripts, then repeat the same workflow in Antigravity CLI if your access tier is affected. Gemini CLI’s public health is excellent. The decision risk is not project abandonment. The decision risk is choosing the wrong Google agent interface for the access tier your team actually uses.
Sources
- https://geminicli.com
- https://github.com/google-gemini/gemini-cli
- https://github.com/google-gemini/gemini-cli/releases/tag/v0.48.0-nightly.20260613.g9e5599c32
- https://github.com/google-gemini/gemini-cli/releases/tag/v0.46.0
- https://github.com/google-gemini/gemini-cli/releases/tag/v0.47.0-preview.0
- https://developers.googleblog.com/an-important-update-transitioning-gemini-cli-to-antigravity-cli