Google's open-source terminal agent with Gemini 3 models, 1M token context, built-in Google Search grounding, and the best free tier in the category (60 req/min, 1K req/day). v0.35.0 (Mar 24) shipped keybinding, policy, and telemetry fixes while the repo hit 98,957 stars and 12,593 forks. Terminal-Bench 2.0: 78.4% (#1). SWE-bench Pro standardized 43.30% (#3). Plan Mode added March 2026. First-pass correctness ~50-60% (Educative.io).
Shipped last month
Gemini CLI moves to #2 in coding-clis
Weekly v0.35.0 releases plus doc-verified 1M context / 1K free requests pushed Gemini ahead of Codex CLI in the coding-clis ranking. · 2026-03-25
Where it wins
98,380 stars (#1 by star count in coding agent category)
Best free tier: 1K req/day with personal Google account — unmatched in category
1M token native context window — largest in category
SWE-bench Verified 80.6% (Gemini 3.1 Pro) — near-parity with Claude Code
Terminal-Bench 2.0: 78.4% (#1) — highest terminal-native task score in category
SWE-bench Pro 43.3% (SEAL) — competitive, gap narrowing with each model release
709 commits/month — very active development
Plan Mode added March 11, 2026 — closes last major feature gap
Native Google Search grounding mid-task — unique among coding CLIs
Open source (Apache 2.0), 446 contributors, extensions system + full MCP client
Where to be skeptical
File deletion incident catalogued in AI Incident Database (#1178) — trust-damaging event
Free tier throttles to Flash model after 10-15 Pro prompts (community-reported)
HN consensus: 'Gemini is really bad with tool calling' (158 pts thread)
No published SWE-bench scores from Google themselves — 76.2% comes from independent testing
50-60% first-pass correctness (Educative.io) — roughly half of Claude Code's ~95%. Requires significantly more iteration
CLI scaffolding acknowledged as behind the model's capabilities, even by Gemini enthusiasts
Editorial verdict
#3 coding CLI — best free entry point and Terminal-Bench leader (78.4% #1). 1K req/day free tier is unmatched, 1M context is the largest. SWE-bench Pro standardized 43.30% is competitive. Plan Mode (March 2026) closes last major feature gap. File deletion incident (AI Incident Database #1178) and 50-60% first-pass correctness (roughly half of Claude Code) are real concerns.
How to get started
Run npx @google/gemini-cli or install globally with npm install -g @google/gemini-cli, then authenticate when prompted (Sign in with Google for the free 1,000 requests/day tier or supply a Gemini Code Assist license). From your repo run gemini plan/gemini todo to let it inspect the project before you approve the generated steps and let gemini run execute them.
Videos
Reviews, tutorials, and comparisons from the community.
Introducing Gemini CLI
Gemini CLI Tutorial #1 - Introduction & Setup
Gemini CLI - 11 PRO TIPS in 25 Minutes
Related
Coding CLIs / Code Agents
Budget-conscious developers, students, exploratory/prototyping work, massive context window tasks
Software Factories
Budget-conscious developers, Google Cloud/Android shops, and anyone wanting a capable free coding CLI
Claude Code
98Anthropic's official agentic coding CLI. v2.1.81 (Mar 20) shipped `--bare`, smarter worktree resume, and improved MCP OAuth while the repo crossed 82,204 stars and logged ~14 commits/week across 10+ maintainers. Terminal-native, tool-use-driven, with deep file system + shell access, #1 SWE-bench Pro standardized (45.89%), ~4% of GitHub public commits (SemiAnalysis), $2.5B annualized revenue. 8M+ npm weekly downloads. Opus 4.6 with 1M context.
OpenHands
88Category leader in multi-agent orchestration — 69,352 stars (verified), $18.8M Series A, AMD hardware partnership, 455 contributors, 1M downloads/month PyPI (3.4M all-time). SWE-Bench Verified 72% with Claude 4.5 Extended Thinking (updated 2026-03-19), Multi-SWE-Bench #1 across 8 languages. Gap to #2 is enormous on every axis.
OpenCode
88Open-source AI coding agent from SST. v1.2.27 active (2026-03-16) — development resumed after a gap. OpenAI official partnership following the Anthropic OAuth block controversy. 126K+ GitHub stars (star surge driven by Anthropic controversy). Known unauthenticated RCE fixed in v1.1.10+ (CVE, 432 HN pts). CVE-2026-22812 (CVSS 8.8-10.0) is a second serious security incident.
CodeQL (via GitHub MCP Server)
88GitHub-native SAST via CodeQL, accessible through the official GitHub MCP Server. Copilot Autofix generates fixes from CodeQL alerts. GitHub Security Lab Taskflow Agent found ~30 real CVEs. Zero extra setup for GitHub users.
Public evidence
Latest release notes confirm weekly shipping: new keybinding overrides, policy tweaks, telemetry fixes, and multi-platform binaries. Confirms the repo’s rapid cadence before we elevated it to #2.
"Gemini CLI is the better starting point ... The 1M context window means you can work on personal projects of any size without worrying about token limits" before recommending a switch to Claude Code when hallucinations creep in.
Table shows Claude Code at 200K tokens for $20/mo, Copilot CLI as paid-only, and Gemini CLI with 1M tokens plus 1,000 requests/day for free. Reinforces why developers start on Gemini even if they later upgrade elsewhere.
Praise for free tier and easy setup. Major criticism: privacy policy (free-tier data used for training), npm packaging. Hit 17K stars within 24 hours.
30 pro-tips for agentic coding. HN discussion was skeptical: multiple commenters reported tool-calling loops, several preferred Claude Code for complex tasks.
Called it 'very good.' Praised million-token context and decision-making taste. Flagged free-tier training data concern. Noted it's open-source unlike Claude Code.
Covers Conductor extension adding automated code reviews, test validation, and security scanning. Shows Gemini CLI evolving into a full development workflow tool.
Scores 86/100 vs Claude Code's 98/100. Excels at terminal-first workflows and free-tier access. Notes Gemini 3 Flash achieves 78% SWE-bench.
Gemini CLI scores 43.30% on SWE-bench Pro standardized, placing #2 behind Claude Code (45.89%) and ahead of Codex CLI (41.04%). The gap narrowed significantly with Gemini 3 Pro — trajectory is strong.
Gemini CLI took 2h02m with multiple retries to complete a React project vs Claude Code's 1h17m single-shot. Practical performance gap is wider than SWE-bench Pro spread suggests.
File deletion pattern is not isolated — cataloged as AI Incident Database entry #1178 with multiple GitHub issues. 'I have failed you completely and catastrophically' (Gemini's own admission). Trust-breaking for production use.
Raw GitHub source
Constrained peek so you can sanity-check the source material without leaving the site.
Gemini CLI is an open-source AI agent that brings the power of Gemini directly into your terminal. It provides lightweight access to Gemini, giving you the most direct path from your prompt to our model.
Learn all about Gemini CLI in our documentation.
See Gemini CLI installation, execution, and releases for recommended system specifications and a detailed installation guide.
# Using npx (no installation required)
npx @google/gemini-cli
npm install -g @google/gemini-cli
brew install gemini-cli
sudo port install gemini-cli
# Create and activate a new environment
conda create -y -n gemini_env -c conda-forge nodejs
conda activate gemini_env
# Install Gemini CLI globally via npm (inside the environment)
npm install -g @google/gemini-cli
See Releases for more details.
New preview releases will be published each week at UTC 23:59 on Tuesdays. These
releases will not have been fully vetted and may contain regressions or other
outstanding issues. Please help us test and install with preview tag.
npm install -g @google/gemini-cli@preview
preview release + any bug fixes
and validations. Use latest tag.npm install -g @google/gemini-cli@latest
nightly tag.npm install -g @google/gemini-cli@nightly
Integrate Gemini CLI directly into your GitHub workflows with Gemini CLI GitHub Action:
@gemini-cli in issues and pull requests
for help with debugging, explanations, or task delegation