Gemini CLI
Google's open-source terminal agent for Gemini, launched in 2025 and now maintained for enterprise Gemini Code Assist workflows. It runs a tool loop over file system operations, shell commands, and web fetch and search, reads project context from GEMINI.md files, connects to MCP servers, and has a headless mode with JSON output for scripts and CI. Google narrowed its role in 2026: free individual access and Google AI Pro and Ultra plans moved to the separate Antigravity CLI on June 18, 2026, and the README now says Gemini CLI is not tuned for the Gemini 4 series and should not be used to benchmark Gemini models. Its Terminal-Bench 2.1 runs use earlier Gemini models. Gemini 3 Pro scores 65.8 in Gemini CLI against 73.9 in the Terminus 2 harness on the same board, while Gemini 3.1 Pro is level across the two at 65.8 and 65.6.
Leaderboard Placements
Live board as of 2026-10-28: federated Terminal-Bench rows, with TensorFeed editorial scores for harnesses the board does not cover.
| Benchmark | Best base model | Score | Rank |
|---|---|---|---|
| SWE-bench Verified | - | - | - |
| Terminal-Bench | - | - | - |
| Terminal-Bench 2.1 | Gemini 3 Pro | 65.8 | #10 / 12 |
| Terminal-Bench 3.0 | - | - | - |
| Terminal-Bench 4.0 | - | - | - |
| Aider Polyglot | - | - | - |
| SWE-Lancer | - | - | - |
Distribution
Open-source CLI under the Apache 2.0 license. Install with npm (@google/gemini-cli), run it with npx, or use Homebrew or MacPorts. A VS Code companion extension is available.
Model Story
Google Gemini models only, chosen per run with the -m flag. Google's README says it is not tuned for the Gemini 4 series.
Pricing
Free, open-source harness with no free model tier. Sign in with Google requires a paid Gemini Code Assist license; otherwise usage bills to a Gemini API key or a Vertex AI project. Free individual access moved to Antigravity CLI on June 18, 2026.
Who It's For
Teams on Gemini Code Assist Standard or Enterprise, and Google Cloud shops that want an open-source, scriptable terminal agent billed through Vertex AI.
Notable Features
- MCP servers configured in ~/.gemini/settings.json
- GEMINI.md project context files
- Headless mode (-p) with json and stream-json output
- Custom extensions for shared commands
- Weekly stable and preview releases plus nightly builds