← Back to research
·8 min read·opensource

Gemini CLI

Gemini CLI is Google's open-source AI coding agent (105K+ GitHub stars), now being sunset for individual users on June 18, 2026 in favor of the closed-source Antigravity CLI. Enterprise access continues via Gemini Code Assist.

Key takeaways

  • Status change: Gemini CLI stops serving free, Google AI Pro, and AI Ultra requests on June 18, 2026 — Google is migrating individuals to the closed-source Antigravity CLI
  • Free tier (through June 18, 2026): 1,000 requests/day with a personal Google account; Gemini 3 models with 1M token context
  • Enterprise path survives: Gemini Code Assist Standard ($19/user/mo) and Enterprise ($45/user/mo) licenses keep Gemini CLI access with the latest Gemini models

FAQ

What is Gemini CLI?

Gemini CLI is Google's open-source AI coding agent that runs in your terminal with access to Gemini 3 models and a 1M token context window.

Is Gemini CLI being shut down?

Partially. On June 18, 2026, Gemini CLI stops serving requests for free, Google AI Pro, and AI Ultra users, who are directed to the closed-source Antigravity CLI. Gemini Code Assist Standard and Enterprise customers keep Gemini CLI access.

Is Gemini CLI free?

Through June 18, 2026, the free tier includes 1,000 requests/day with a personal Google account. After that date, individual access moves to Antigravity CLI; only Code Assist Standard/Enterprise licenses retain Gemini CLI.

What's special about Gemini CLI's context window?

Gemini CLI has a 1M token context window — significantly larger than Claude Code or Codex.

Does Gemini CLI compete with Tembo?

Gemini CLI is a coding agent; Tembo orchestrates coding agents. They're complementary layers.

Executive Summary

Gemini CLI is Google's official open-source AI coding agent — 105K+ GitHub stars as of June 2026 — known for the most generous free tier among major coding agents (1,000 requests per day with a personal Google account) and a 1M token context window.[1][2]

The headline as of June 2026 is the sunset: announced at Google I/O on May 19, 2026, Gemini CLI and the Gemini Code Assist IDE extensions stop serving requests for free, Google AI Pro, and Google AI Ultra users on June 18, 2026. Individuals are directed to the closed-source Antigravity CLI; Gemini Code Assist Standard and Enterprise customers keep Gemini CLI access with the latest Gemini models.[3][4]

AttributeValue
CompanyGoogle
Founded1998 (Gemini CLI: 2025)
FundingPublic company (GOOGL)
Employees~190,000
HeadquartersMountain View, CA

Product Overview

Gemini CLI is Google's official open-source AI coding agent (Apache 2.0), with 105K+ stars, 14K forks, and thousands of merged community pull requests as of June 2026.[1] It brings Gemini 3 models directly into the terminal, and also powers agent mode inside the Gemini Code Assist IDE extensions for VS Code and JetBrains.[4]

The standout feature is the 1M token context window — enough to load entire codebases that would overflow other agents. Combined with built-in Google Search grounding, Gemini CLI can answer questions with real-time information.[5]

Transition notice: On June 18, 2026, Gemini CLI stops serving requests for free, Google AI Pro, and AI Ultra tiers. Google's replacement, Antigravity CLI, is a Go-based, multi-agent, closed-source tool with weekly (not daily) quotas. Enterprise access via Gemini Code Assist Standard/Enterprise licenses continues unchanged.[3][6]

Key Capabilities

CapabilityDescription
1M Token ContextLargest context window among major agents
Google Search GroundingReal-time information in responses
MultimodalGenerate code from images, PDFs, sketches
MCP SupportModel Context Protocol for custom integrations
ExtensionsPackaged extension ecosystem (commands + MCP servers)
GitHub ActionNative CI/CD integration

Product Surfaces / Editions

SurfaceDescriptionAvailability
CLITerminal-based agentGA (individual tiers until June 18, 2026)
Gemini Code Assist IDEAgent mode in VS Code / JetBrainsGA (Standard/Enterprise continue)
GitHub ActionCI/CD automationGA
Non-Interactive ModeScripting integrationGA

Technical Architecture

Installation:[1]

# Quick start (no installation)
npx @google/gemini-cli

# Global install
npm install -g @google/gemini-cli
brew install gemini-cli  # macOS/Linux

Key Technical Details

AspectDetail
DeploymentLocal CLI
Model(s)Gemini 3 model family (1M context)
IntegrationsGoogle Search, GitHub, MCP, extensions
Open SourceYes (Apache 2.0) — successor Antigravity CLI is closed source

Authentication options:

  1. Google OAuth — Free tier with personal account (sunsets June 18, 2026) [7][8]
  2. Gemini API Key — Usage-based billing; unpaid keys limited to 250 requests/day on Flash models[2]
  3. Vertex AI / Code Assist license — Enterprise features, Google Cloud integration; this path continues post-transition[4]

Strengths

  • Generous free tier — 1,000 requests/day with no payment required (through June 18, 2026)[2]
  • 1M token context — Largest context window among major agents
  • Google Search grounding — Real-time information in responses[5]
  • Multimodal — Generate code from images, PDFs, sketches
  • MCP + extensions — Extensible with custom integrations
  • GitHub Action — Native CI/CD integration[9]
  • Massive community — 105K+ stars, 14K forks, thousands of merged PRs[1]
  • Enterprise continuity — Code Assist Standard/Enterprise customers keep supported access[3]

Cautions

  • Individual access ends June 18, 2026 — Free, AI Pro, and AI Ultra tiers stop being served; migration to Antigravity CLI required[3]
  • Open-source trust damage — Developers on Hacker News called the move a bait-and-switch: thousands of community PRs merged into an Apache 2.0 project, then core development shifted to the closed-source Antigravity CLI[10][6]
  • Successor quota regression — Antigravity CLI uses weekly (not daily) quotas; early users reported exhausting them in a handful of requests[6]
  • Google-only — Locked to Gemini models
  • Google ecosystem dependence — Tied to Google account/Cloud; the license protected the code but not the backend that serves requests

Pricing & Licensing

TierPriceIncludes
Free (until June 18, 2026)$01,000 req/day, Gemini 3, 1M context[2]
Google AI ProSubscription1,500 req/day[2]
Google AI UltraSubscription2,000 req/day[2]
Gemini API key (unpaid)$0250 req/day, Flash models only[2]
Gemini API key (paid)Pay-per-useUsage-based billing
Code Assist Standard$19/user/mo (annual; $22.80 monthly)1,500 req/day, Gemini CLI continues post-June 2026[11][4]
Code Assist Enterprise$45/user/mo (annual; $54 monthly)2,000 req/day, code customization, GCP integration[11][4]

Licensing model: Open source (Apache 2.0) + paid API / Code Assist seats; successor Antigravity CLI is closed source[6]

Hidden costs: None for free tier while it lasts; Vertex AI requires Google Cloud account[1]


Competitive Positioning

Direct Competitors

CompetitorDifferentiation
Claude CodeClaude Code has 200K context; Gemini CLI has 1M — but Claude Code's individual tiers aren't being sunset
Codex CLICodex requires ChatGPT plans; Gemini CLI was free (until June 18, 2026)
Antigravity CLIGoogle's own closed-source, multi-agent successor for individual users
OpenCodeOpenCode is model-agnostic; Gemini CLI is Google-only but simpler
CursorCursor is full IDE; Gemini CLI is terminal-first

When to Choose Gemini CLI Over Alternatives

  • Choose Gemini CLI when: You hold a Code Assist Standard/Enterprise license, or need 1M context and Google Search grounding before the individual sunset
  • Choose Claude Code when: You prefer Anthropic models or want a stable individual subscription path
  • Choose Codex when: You want the OpenAI ecosystem experience
  • Choose OpenCode when: You need model flexibility across providers

Ideal Customer Profile

Best fit:

  • Organizations with Gemini Code Assist Standard/Enterprise licenses (access continues)
  • Teams working with large codebases needing 1M token context
  • Engineers who value Google Search grounding for real-time information
  • GitHub users wanting native CI/CD integration

Poor fit:

  • Individual developers planning beyond June 18, 2026 (forced migration to Antigravity CLI)
  • Developers who require an open-source tool with a guaranteed open roadmap
  • Organizations requiring model flexibility

Viability Assessment

FactorAssessment
Financial HealthStrong — Google has unlimited resources
Market PositionTransitioning — individual tiers sunset June 18, 2026 in favor of Antigravity CLI
Innovation PaceRedirected — core agent investment moving to Antigravity
Community/EcosystemLarge but bruised — 105K+ stars, with open-source contributors publicly criticizing the closed-source pivot[10]
Long-term OutlookEnterprise-only continuity via Gemini Code Assist; individual product effectively deprecated

Google has the resources to sustain Gemini CLI indefinitely — but has chosen not to for individuals. The deprioritization risk flagged in earlier versions of this profile materialized: Google's agent strategy shifted to Antigravity, and Gemini CLI's individual tiers are being switched off on June 18, 2026.[3]


Bottom Line

Gemini CLI was the most accessible entry point to AI coding agents — 1,000 free requests per day and a 1M token context window earned it 105K+ GitHub stars. That era ends June 18, 2026, when Google stops serving free, AI Pro, and AI Ultra requests and pushes individuals to the closed-source Antigravity CLI.[3]

The enterprise path is the survivor: Gemini Code Assist Standard ($19/user/mo) and Enterprise ($45/user/mo) licenses keep Gemini CLI supported with the latest Gemini models.[11]

Recommended for: Organizations on Gemini Code Assist Standard/Enterprise who want a supported, Gemini-3-powered terminal agent with 1M context.

Not recommended for: Individual developers — the free tier disappears June 18, 2026, and the open-source community's trust in the successor is low.

Outlook: Gemini CLI becomes an enterprise-licensed tool; Google's individual coding-agent energy moves to Antigravity. Watch whether Antigravity CLI's weekly quotas and closed-source posture push free-tier refugees toward Claude Code, Codex, or open alternatives.[6]


Research by Ry Walker Research • methodology