⌂ Home
Gemini Code Assist: AI Pair Programming in Your IDE
Code Assist Track
⏰ 25-35 min Module 1 of 12 📚 No code required
🧱
Gemini Code Assist Track — G00: Orientation Before you install anything, understand what Gemini Code Assist is, what it can do, which edition you need, and how the 12 modules of this course fit together. Zero code in this module — just the map.

Orientation: What Is Gemini Code Assist?

You're about to spend 12 modules learning a tool that lives inside your editor and changes how you write software. This module gives you the complete picture first, so every later module slots into a map you already have.

What Is Gemini Code Assist?

Analogy First

Imagine being assigned a new teammate who has read every programming book ever published, every page of documentation for every library you use, and millions of open-source repositories. They sit next to you all day. When you start typing a function, they quietly suggest how it might continue. When you're stuck on an error, you turn to them and ask. When you have a tedious multi-file change, you hand them the whole task and review their work before it lands. Before this teammate existed, all of that knowledge was scattered across Stack Overflow tabs, docs sites, and your own memory — and every lookup broke your flow. Gemini Code Assist is that teammate, installed as an extension in the editor you already use. It maps onto your daily work in three modes: autocomplete-style suggestions while you type, a chat panel when you want to ask, and an agent mode when you want to delegate.

Technical Definition

Gemini Code Assist is Google's AI coding assistant: a set of IDEIntegrated Development Environment — the application you write code in, such as Visual Studio Code, IntelliJ IDEA, PyCharm, or Android Studio. extensions, a command-line tool, and a GitHub app, all powered by Google's Gemini modelsGoogle's family of large language models. Gemini 2.5 Flash and Pro are generally available in Code Assist; Gemini 3.x models are rolling out in preview.. It provides code completionReal-time, inline suggestions that appear as faded "ghost text" while you type. Press Tab to accept., code generation from natural-language prompts, a chat assistant that understands your open files, an agent modeA mode where the AI plans and executes multi-step, multi-file tasks on its own, showing you diffs to approve before changes are applied. for multi-step tasks, and automated code review on pull requests.

It works in VS Code, JetBrains IDEs (IntelliJ, PyCharm, GoLand, WebStorm…), Android Studio, and on GitHub. A companion terminal tool, Gemini CLI, shares the same license and configuration.

Where it sits in your day

The easiest way to understand Gemini Code Assist is to watch one developer task flow through it. Press play:

A day with Gemini Code Assist — VS Code
9:02You type: def convert_grams_to_
      ounces(grams: float) -> float: ← ghost text appears
✓ You press Tab. Completion accepted. (Feature: code completion)
10:15A test fails. You select the error, right-click → "Gemini: Fix"
✦ Chat explains the bug and proposes a one-line fix (Feature: smart actions)
13:40You ask chat: "How does auth work in this project?"
Gemini answers using your actual files as context (Feature: chat + codebase awareness)
15:30You switch to Agent mode: "Add a rating field to recipes — model, API, tests"
Agent plans: 4 files to change. Shows diffs. You review and approve.
✓ Multi-file change applied (Feature: agent mode)
16:45You open a pull request on GitHub
✦ Gemini Code Assist posts a PR summary + line-by-line review (Feature: GitHub app)
Press ▶ to watch a full day
Why It Matters

Google's own studies and DORA research report that developers using AI assistance complete common tasks measurably faster — and the biggest wins are not the autocomplete keystrokes. They come from the expensive interruptions that disappear: the 20-minute doc search that becomes a 20-second chat answer, the half-day multi-file refactor that becomes a reviewed agent run. If you write code for even 2 hours a day, mastering this tool pays for the ~8 hours this course takes within your first couple of weeks.

The Feature Map

Everything Gemini Code Assist does falls into six feature families. Each family gets one or more dedicated modules in this course. Hover over each card — this is the map we'll fill in module by module.

⚡ Code Completion & Generation

Ghost-text suggestions as you type; whole functions generated from a comment.

IDE editor pane • Module G02

💬 Chat, Smart Commands & Actions

A conversation panel that knows your code. Slash commands like /fix and /doc; right-click actions on selected code.

IDE chat panel • Modules G03–G04

🧠 Context & Customization

A 1M-token context window, full local codebase awareness, GEMINI.md rule files, and .aiexclude to keep secrets out.

Project files • Module G05

🤖 Agent Mode

Delegate multi-step, multi-file tasks. Plan review, diff approval, checkpoints, MCP tool servers, auto-approve mode.

IDE chat panel (Agent toggle) • Modules G06–G07

⌨️ Gemini CLI & GitHub Reviews

The same brain in your terminal, plus an app that summarizes and reviews every pull request.

Terminal + github.com • Modules G08–G09

☁️ Google Cloud & Enterprise

Deploy to Cloud Run from chat, Gemini inside BigQuery and Firebase, code customization from your private repos, enterprise security.

Google Cloud console + IDE • Modules G10–G11

One lifecycle, one tool

Press play to see how the features chain across the software lifecycle — this is the same path your capstone project will take in G11:

Writecompletion
Understandchat
Refactor & Testsmart actions
Build Featuresagent mode
ReviewGitHub app
DeployCloud Run
Press ▶ to light up the lifecycle
What Just Happened?

You now have the whole map: six feature families covering write → understand → refactor → build → review → deploy. Every module from here on zooms into exactly one region of this map. If you ever feel lost later, come back to this diagram.

Editions & Pricing

Analogy First

Think of a gym with three membership levels. The free day-pass gets you onto the main floor — real equipment, real workouts, but you can't book the private studios. The Standard membership adds priority booking and bigger usage limits. The Premium membership adds a personal trainer who has studied your training history specifically. Software teams face the same decision pain — paying for capacity they don't need, or hitting limits they didn't expect. Gemini Code Assist's three editions map exactly: free individual tier (the day-pass), Standard (higher limits + enterprise security), and Enterprise (everything, plus a model that has studied your private code).

Individuals

$0 / month
  • Code completion, generation & chat
  • Agent mode & Gemini CLI (generous free quotas)
  • Sign in with a personal Google account
  • No Google Cloud project needed
  • Migrating to Antigravity June 18, 2026 (next section)

Standard

~$19–22.80 / user / month
  • Everything in Individuals, higher quotas
  • Enterprise-grade security & data governance
  • Local codebase awareness (large context)
  • Google Cloud integrations (Cloud Run, BigQuery, Firebase…)
  • IP indemnification
  • Managed via a Google Cloud project

Enterprise

~$45–54 / user / month
  • Everything in Standard
  • Code customization: suggestions tuned on your private repos
  • GitHub / GitLab / Bitbucket code review integration
  • Apigee & Application Integration support
  • Advanced compliance controls
Gotcha

Prices above are list prices as of mid-2026 and vary by region, annual-commit discounts (~17% off), and promotions. Always check the official pricing page before budgeting: cloud.google.com/products/gemini/pricing. Quotas (requests/day, CLI invocations) also differ per edition — see the quotas page.

Why It Matters

Choosing wrong costs real money or real capability. A 10-person team on Enterprise that never uses code customization or PR reviews overpays roughly $3,000/year versus Standard. A solo consultant on the free tier who builds their workflow on it without reading the migration notice (next section) loses their assistant mid-project. Five minutes of edition-mapping now prevents both.

The June 2026 Transition: Antigravity

⚠ Read This Before You Commit To A Workflow

Google has announced that Gemini Code Assist IDE extensions and Gemini CLI stop serving requests for the Individuals (free) tier and Google AI Pro/Ultra subscriber tiers on June 18, 2026. Individual users are directed to migrate to Google Antigravity — Google's newer, agent-first development platform — and Antigravity CLI.

Standard and Enterprise editions are not affected — they continue as the supported path for teams and businesses, which is why this course teaches Gemini Code Assist with the Standard/Enterprise feature set as its backbone. Source: official release notes.

What this means for you
  • You're learning for work / your team has Standard or Enterprise: nothing changes. Everything in this course applies.
  • You're an individual on the free tier: you can still follow every lab today, and your skills transfer — Antigravity uses the same Gemini models, the same agentic concepts (plans, diffs, tool approvals), and very similar context-file conventions. Concepts in G02–G07 map almost one-to-one.
  • You're evaluating tools for a company: factor the roadmap in. Google is consolidating individual-developer AI tooling under Antigravity while keeping Gemini Code Assist as the Google Cloud team product.
What Just Happened?

You learned the single most important "gotcha" in the 2026 Gemini ecosystem before investing hours of lab time. Tool landscapes shift; the durable skills are the workflows (context files, plan→diff→approve, AI code review), and those are exactly what this course drills.

The Models Underneath

Analogy First

A delivery company doesn't send a semi-truck to deliver an envelope, and doesn't send a bicycle courier to move a piano. Before you know what vehicles exist, every delivery feels the same and you can't reason about speed or cost. Gemini Code Assist routes your requests to different Gemini models the same way: a fast, inexpensive model handles the envelope-sized jobs (completions as you type), while a larger reasoning model handles the piano-sized jobs (agent-mode refactors). Knowing the fleet helps you predict behavior — why completions feel instant but agent plans take seconds.

Technical Definition

As of mid-2026, Gemini Code Assist is powered by the Gemini 2.5 family (Flash and Pro, generally available), with Gemini 3-series models (3.0 Flash, 3.1 Pro, 3.5 Flash) rolling out through preview and GA release channels. Two numbers matter most:

  • Context window: up to 1 million tokens — roughly 30,000–50,000 lines of code can be "in the model's head" at once. This is what makes whole-project awareness possible (G05).
  • Release channels: admins can pin teams to GA-only features or opt into preview features — useful for enterprises that need stability.
Why It Matters

A 1M-token context window means the difference between an assistant that sees one file and one that sees your whole service. Concretely: asked "where is this discount actually applied?", a single-file assistant guesses; a codebase-aware assistant answers with the three call sites across pricing.py, checkout.py, and promotions.py. You'll prove this to yourself in the G05 lab.

How This Course Works

Twelve modules, roughly 70% hands-on. From G02 onward you build and evolve one project — Recipe Box, a small recipe-management web API — so each module's lab starts where the last one ended. By G11 you'll have written, tested, reviewed, and deployed it with Gemini Code Assist doing the heavy lifting at every stage.

ModuleTopicYou build
G00Orientation (you are here)The mental map
G01Setup & first suggestionWorking install in VS Code / JetBrains
G02Code completion & generationRecipe Box core from comments alone
G03Chat, smart commands & actionsDebugged + documented endpoints
G04Refactoring, tests & debuggingA test suite and cleaned-up legacy code
G05Context: GEMINI.md & .aiexcludeProject rules that change suggestions
G06Agent mode I: multi-file tasksA feature spanning 5 files
G07Agent mode II: MCP & toolsA GitHub-connected agent workflow
G08Gemini CLITerminal automation for Recipe Box
G09AI code reviews on GitHubAuto-reviewed pull requests
G10Google Cloud integrationsRecipe Box live on Cloud Run
G11Enterprise, security & capstoneThe full lifecycle, end to end

What you need

  • A Google account (free tier works for most labs through June 17, 2026; a Standard license or trial is the durable path)
  • VS Code (primary in labs) or a JetBrains IDE — both are covered in G01
  • Python 3.10+ installed (Recipe Box is a small Flask API; every snippet is complete and runnable)
  • A GitHub account (for G07 and G09)
  • A Google Cloud account with billing only for G10's deploy lab (free-tier credits cover it; the module includes a no-cloud fallback)
What Just Happened?

You've seen the destination (a deployed, AI-reviewed Recipe Box) and the route (12 modules across 6 feature families). Time to verify you absorbed the map — then on to G01 to install the tool.

Knowledge Check

1. Which three "modes" describe how Gemini Code Assist fits into daily coding work?

Compile, debug, deploy
Suggest while you type, answer when you ask, act when you delegate
Lint, format, refactor
Search, copy, paste

2. A 10-person team wants higher quotas, enterprise security, and Cloud Run integration — but doesn't need suggestions tuned on its private repos. Which edition fits?

Individuals (free)
Standard
Enterprise
They must build their own with the Gemini API

3. What happens to Gemini Code Assist on June 18, 2026?

The entire product shuts down for all users
It becomes free for everyone
The Individuals (free) and AI Pro/Ultra tiers stop being served; those users migrate to Antigravity. Standard/Enterprise continue.
Only the JetBrains plugin is discontinued

4. Which capability is exclusive to the Enterprise edition?

Agent mode
Code customization — suggestions informed by your organization's private repositories
Code completion
Using GEMINI.md context files

5. Why does the 1-million-token context window matter in practice?

It makes individual completions appear faster
It reduces the subscription price
The model can hold your whole project in context, so answers reference real code across many files instead of guessing from one file
It allows offline use without an internet connection

Quiz Complete!

Ready for G01: Setup & Your First Suggestion →