🔮 Gemini CLI Track

Building with
Gemini CLI

From first prompt to production workflows — 18 modules covering setup, SDLC, MCP integration, Google Cloud, GitHub Actions, and Google Workspace automation.

🔍 Search the Course  ·  💬 Ask the Course (works offline)

18Modules
$0Free Tier
1MToken Context
~38hLearning Time
What is Gemini CLI?

Gemini CLI is Google's open-source AI agent for your terminal. It brings Gemini 2.5 Pro/Flash directly to your shell — reading files, running commands, calling MCP servers, and automating your entire development workflow. Unlike web chat, it operates in your project context: it can see your code, run tests, commit to Git, and trigger CI pipelines. The free tier includes 60 requests per minute and 1,000 requests per day using Gemini 2.5 Pro — the most capable model available in any free AI CLI tool.

1
Type a prompt
In your terminal, in any directory. Gemini sees your files via @file mentions.
2
Gemini plans
Plan Mode shows what it will do before touching anything. You approve or edit the plan.
3
Tools execute
Built-in tools read/write files, run shell commands, call MCP servers, search the web.
4
You review
Diffs shown inline. Memory persists the context for your next session or CI run.
Quick Setup
# Windows (PowerShell — run as user, NOT admin) Set-ExecutionPolicy RemoteSigned -Scope CurrentUser winget install OpenJS.NodeJS.LTS npm install -g @google/gemini-cli gemini # opens Google OAuth in browser — sign in with your Google account
# Mac / Linux npm install -g @google/gemini-cli gemini # → opens Google OAuth in your default browser # → sign in → terminal shows: ✓ Authenticated as you@gmail.com
No API key required for the free tier. Google OAuth authenticates you with your personal Google account — the same account you use for Gmail and Google Drive. The free tier gives you 60 requests/min and 1,000 requests/day on Gemini 2.5 Pro. For production workloads, set GEMINI_API_KEY in your environment to switch to the paid API.
Track Modules — Learning Path
TRACK 1 — GETTING STARTED
MODULE 01 · GETTING STARTED

Dev Setup & First Run

Windows-first setup: Node.js via WinGet, execution policy fix, global npm install. Google OAuth login flow. First commands: file analysis, code generation, shell execution. Free tier limits and rate-limit behavior.

LIVE ~50 min Windows-first
MODULE 02 · GETTING STARTED

Prompting & Commands

@file and @folder mentions to load context. Built-in /slash commands reference. Memory system (/memory add, /memory show). Multiline input with triple-backtick blocks. Effective prompting patterns for code tasks.

~55 min Beginner Core Skills
MODULE 03 · GETTING STARTED

GEMINI.md & Configuration

The project context file that loads every session. Model selection (Flash vs Pro), temperature, system instructions, token caching. Global vs project-level config. Three starter templates: API project, monorepo, data pipeline.

~55 min Beginner → Intermediate Config
TRACK 2 — SKILLS & AGENTIC WORKFLOWS
MODULE 04 · AGENTIC WORKFLOWS

skill.md & Custom Commands

The .gemini/skills/ directory. Writing reusable skill.md files with argument placeholders. Sharing skills across projects. Building a personal skill library: PR summarizer, test generator, release notes writer.

~65 min Intermediate .gemini/skills/
MODULE 05 · AGENTIC WORKFLOWS

Plan Mode

Planning vs execution separation: use Plan Mode before destructive or multi-step operations. How Gemini routes to the best model for planning. Approval gates: inspect, edit, and approve plans before they execute. Integrating Plan Mode into team workflows.

~55 min Intermediate Approval Gates
TRACK 3 — AI SDLC
MODULE 06 · AI SDLC

Requirements & Planning

Use Gemini CLI as a requirements critic: PRD generation, user story decomposition, INVEST scoring. Feed existing docs via @file for gap analysis. Output machine-readable specs for downstream modules.

~60 min Intermediate PRD · User Stories
MODULE 07 · AI SDLC

Design & Architecture

System design conversations with full codebase context. Mermaid diagram generation (sequence, ER, C4). API contract drafting with OpenAPI 3.1. Database schema design with migration scripts. Architecture decision records (ADR) from a prompt.

~70 min Intermediate Mermaid · OpenAPI
MODULE 08 · AI SDLC

Implementation

Scaffold full projects from a spec file. Multi-file context: load entire src/ trees, generate code that fits the existing patterns. Commit message generation from staged diffs. Branch-per-feature workflow with Gemini as co-author.

~75 min Intermediate Scaffolding · Git
MODULE 09 · AI SDLC

Testing & QA

Generate unit tests from implementation files: happy path, edge cases, error branches. Feed test failure output back to Gemini for diagnosis. Coverage gap analysis from lcov reports. Mutation testing prompts to verify test quality.

~65 min Intermediate Pytest · Jest · Coverage
MODULE 10 · AI SDLC

Deployment & Maintenance

Generate Dockerfiles, multi-stage builds, and docker-compose from project structure. CI YAML generation for GitHub Actions and Cloud Build. Runbook authoring from architecture docs. Log analysis: paste error output, get root cause and fix.

~70 min Intermediate → Advanced Docker · CI/CD
TRACK 4 — INTEGRATIONS
MODULE 11 · INTEGRATIONS

MCP Servers

Gemini CLI's native MCP support: add any MCP server in GEMINI.md. FastMCP server development. OAuth 2.0 for remote servers. Google's official partner extensions (Maps, YouTube, Search). Build a custom MCP server and connect it to Gemini in one session.

~75 min Intermediate MCP · FastMCP
MODULE 12 · INTEGRATIONS

Extensions System

Install community extensions from geminicli.com. Write your own extension: manifest, capabilities, distribution. Publish to the gallery. Differences between extensions and MCP servers: when to use which. Extension sandboxing and permissions.

~60 min Intermediate Extensions · geminicli.com
MODULE 13 · INTEGRATIONS

GitHub Actions

Automate PR review with the @gemini-cli trigger in PR comments. Issue triage and labeling workflows. Action YAML: headless mode, output parsing, conditional steps. Security: OIDC auth instead of long-lived tokens.

~70 min Intermediate → Advanced GitHub Actions · CI
MODULE 14 · INTEGRATIONS

Google Workspace

100+ agent skills for Gmail, Google Docs, Drive, and Sheets via the official Workspace extension. OAuth flow for Workspace APIs. Real workflows: summarize email threads, generate Docs from specs, update Sheets from code output, organize Drive folders.

~65 min Intermediate Gmail · Sheets · Drive
MODULE 15 · INTEGRATIONS

Google Cloud MCP Servers

Official MCP servers for BigQuery, Cloud Storage, Cloud Run, Pub/Sub, and Spanner. Natural language → SQL on your data warehouse. ADC authentication, minimum-privilege IAM setup, and compound workflows that orchestrate multiple Google Cloud services in a single Gemini prompt.

~75 min Intermediate → Advanced BigQuery · Cloud Run · GCS
TRACK 5 — POWER USAGE
MODULE 16 · POWER USAGE

Headless & CI-CD

Non-interactive mode with the gemini -p "prompt" flag. --output-format json for machine-readable output. --yolo for fully unattended runs. Full pipeline integration: trigger from GitHub Actions, parse JSON responses, branch on results.

~70 min Advanced Headless · JSON output
MODULE 17 · POWER USAGE

Memory & Checkpoints

/memory add for persistent session facts. /restore to resume from a saved checkpoint. Global vs project memory scope. Branching memory: keep multiple working contexts for different features. Memory export for handoff to teammates.

~60 min Advanced Memory · /restore
MODULE 18 · POWER USAGE

Shell Automation

Combine Gemini CLI with PowerShell and bash scripts. --sandbox flag with Docker for safe code execution. Automation scripts: daily standup generator, weekly PR digest, release pipeline assistant. Cross-platform script patterns (Windows + Linux parity).

~65 min Advanced PowerShell · bash · Docker
Capstone Project
CAPSTONE · After Module 17 + 18

Full SDLC Build — spec to deployed PR, entirely in Gemini CLI

Start with a one-paragraph product brief. Use Gemini CLI to: generate a PRD, decompose into user stories, produce a system design with Mermaid diagrams, scaffold the implementation, generate and run tests, write the Dockerfile and CI YAML, open a PR, and trigger automated PR review — all from your terminal. Both Windows PowerShell and bash paths shown side-by-side.

5–7 hours All 5 Tracks Windows + bash
Appendix
APPENDIX — WINDOWS DEEP REFERENCE

Windows Deep Reference

Everything Windows developers need: execution policy explained (RemoteSigned vs Bypass vs Unrestricted), WinGet vs chocolatey vs manual Node.js install, Windows Terminal profiles for color-coded Gemini sessions, WSL2 setup for Linux parity, PATH troubleshooting for global npm packages, Docker Desktop configuration for the --sandbox flag, and a lookup table of the 15 most common Windows-specific errors with fixes.

Reference Windows 10/11 WSL2
How Gemini CLI Compares

Not all AI CLIs are the same. Here's what's materially different.

🔮 Gemini CLI
Context Window
1M tokens (Gemini 2.5 Pro)
Free Tier
60 req/min · 1,000 req/day
Google Workspace
Native — Gmail, Docs, Sheets, Drive
Plan Mode
Built-in — inspect before execute
MCP Support
Native via GEMINI.md
Open Source
Yes (Apache 2.0)
Auth Model
Google OAuth (no API key needed)
⚡ Claude Code
Context Window
200K tokens (Claude 3.5)
Free Tier
Paid only (Anthropic API key)
Google Workspace
Via MCP only
Plan Mode
Thinking mode — not same as plan gate
MCP Support
Native + full hooks system
Open Source
No
Auth Model
API key
🐙 GitHub Copilot CLI
Context Window
Varies by backend (GPT-4o)
Free Tier
2,000 completions/mo (limited)
Google Workspace
No
Plan Mode
No
MCP Support
No native support
Open Source
No
Auth Model
GitHub account
Bottom line: Gemini CLI wins on context window (5× Claude Code, 10× most tools), free tier generosity, and native Google Workspace integration. Claude Code wins on the MCP hooks ecosystem and enterprise-grade access controls. GitHub Copilot CLI wins for teams already invested in the GitHub platform. All three support similar shell-integrated code workflows.

Prerequisites

This track is beginner-friendly from Module 01. For Track 4 (MCP Servers) and the Capstone, basic shell comfort and Node.js familiarity are helpful. You do not need to know Python or the Gemini API to start.

Node.js 18+ — for npm install
Google Account — for OAuth login
Terminal comfort — cd, ls, basic git
Windows / Mac / Linux — all supported

Related Tracks & Resources

Main Course (Claude API) 30 modules using the Anthropic API. Deep coverage of agent architecture, RAG, MCP, and production deployment. MCP Track 8 modules on building MCP servers and clients. The MCP servers you build here connect directly to Gemini CLI via GEMINI.md. Open Source Track (Mistral) 20 modules using Ollama and local models. Parallel SDLC patterns to what you learn here, fully offline.