Gemini Code Assist vs GitHub Copilot

Gemini Code Assist and GitHub Copilot do the same everyday job: completions as you type, a chat panel that knows your open files, an agent mode that edits across files and runs commands, and a reviewer for pull requests. The real difference is the platform behind each. Code Assist is bought, administered and governed through Google Cloud, with Google identities, IAM roles and VPC Service Controls. Copilot is bought and governed through GitHub, with policies set by organisation and enterprise owners, and it reaches further into GitHub itself: a cloud agent, a CLI and code review on every pull request. Copilot also supports more editors and offers models from several vendors. So the question for most teams is less “which assistant is smarter” and more “which platform do we already run on”.

This page compares the two for teams. If you are weighing Code Assist against Google’s terminal agent instead, see Gemini CLI vs Gemini Code Assist.

The short version

• Buying: Code Assist Standard and Enterprise through Google Cloud. Copilot Business and Enterprise through GitHub, with individual plans alongside.
• Editors: Code Assist in VS Code, JetBrains IDEs, Android Studio, Cloud Workstations and Cloud Shell Editor. Copilot in VS Code, Visual Studio, JetBrains IDEs, Eclipse and Xcode, with completions in Neovim.
• Agent mode: both, with MCP servers. Google marks its agent mode as Preview.
• Pull request review: Copilot code review is built into GitHub. Google’s reviewer is a separate GitHub app, installed through Google Cloud.
• Models: Code Assist uses Gemini. Copilot offers a choice of models from several providers.
• Governance: Google Cloud IAM, VPC Service Controls and Cloud Logging, or GitHub policies and content exclusion.

Editions, by name

Google sells two editions to organisations. Its Code Assist overview↗ (opens in a new tab) describes Standard as AI coding assistance with enterprise-grade security, and Enterprise as everything in Standard plus customisation from your private repositories and assistance across more Google Cloud services, such as Apigee and Application Integration. The consumer edition, Gemini Code Assist for individuals, stopped serving requests on 18 June 2026, along with sign-ins through Google AI Pro and Ultra, and Google now points those users to Antigravity.

GitHub lists more plans. GitHub’s plans page names Copilot Free, Student, Pro, Pro+ and Max for individuals, and Copilot Business and Enterprise for organisations, with usage counted in GitHub AI credits. Enterprise requires GitHub Enterprise Cloud, and Copilot is not available on GitHub Enterprise Server. Check each vendor’s current pricing for costs and allowances.

Editors and everyday assistance

In the editor the two feel alike: inline completions, whole functions from a comment, a chat panel that works with your open files, and right-click or slash actions such as explaining code or writing tests. Code Assist adds source citations when it quotes at length from existing code. Copilot has an equivalent in code referencing.

Editor coverage is where they part. GitHub’s feature matrix↗ (opens in a new tab) shows agent mode, chat and MCP in VS Code, Visual Studio, JetBrains IDEs, Eclipse and Xcode. If part of your team works in Visual Studio, Eclipse or Xcode, Copilot covers them and Code Assist does not. If they work in Android Studio or Google’s Cloud Workstations, Code Assist is built for that.

Agent mode and MCP

Both have an agent mode: a chat that plans a multi-step task, reads and writes files, runs terminal commands and calls MCP tools, asking for your approval as it goes. Google’s agent mode guide↗ (opens in a new tab) marks it as Preview and adds two cautions: there is no undo for changes the agent makes outside your IDE, and it does not cite sources the way standard chat does. In VS Code it is powered by Gemini CLI and reads MCP servers from ~/.gemini/settings.json, with coreTools and excludeTools settings to allow or block particular tools and commands.

Copilot’s agent mode is generally available across its main editors, and it has checkpoints to roll back a session’s edits. MCP reaches every Copilot surface: the IDE, Copilot CLI, the Copilot app and the cloud agent. There is one admin detail to know. According to GitHub’s MCP documentation↗ (opens in a new tab), the MCP servers in Copilot policy is off by default for Business and Enterprise, so an organisation owner must turn it on before anyone can add a server.

Code review

Copilot code review is part of GitHub. You request it on a pull request like any reviewer, or turn on automatic review for a repository, and it also runs in VS Code, Visual Studio, JetBrains IDEs and Xcode. On Business and Enterprise, owners can let members without a Copilot licence request reviews on GitHub.com. Its suggestions can be handed to the Copilot cloud agent, which opens a pull request with the fixes.

Google’s reviewer is Gemini Code Assist on GitHub. It summarises pull requests, reviews them, and answers /gemini commands in comments. Google’s review documentation↗ (opens in a new tab) says the enterprise version is in Preview, is a separate product from Code Assist Enterprise, connects through Google Cloud’s Developer Connect, supports GitHub Enterprise Cloud and Enterprise Server, and skips files in .github/workflows. The consumer version of the GitHub reviewer is deprecated along with the individuals edition.

Context and customisation

• Instruction files: Code Assist agent mode reads GEMINI.md, and in IntelliJ also AGENT.md. Copilot reads .github/copilot-instructions.md, path-specific .instructions.md files, and AGENTS.md, CLAUDE.md or GEMINI.md, with organisation-wide instructions on GitHub.com.
• Your own code: Code Assist Enterprise’s code customisation indexes private repositories in GitHub, GitLab or Bitbucket, so suggestions follow your organisation’s code. Both products also index the local workspace.
• Models: Code Assist runs Gemini models. Copilot lets users pick from models by Anthropic, OpenAI, Google and others, as far as the plan and the organisation’s policy allow.

Admin and data controls

Code Assist inherits Google Cloud’s controls. Developers sign in with accounts managed by Cloud Identity, Google Workspace or a federated identity provider, and access is granted by IAM roles, ideally to a group. Google’s security, privacy and compliance page↗ (opens in a new tab) adds VPC Service Controls perimeters, optional logging of prompts and responses to Cloud Logging, and a commitment not to use your data to train models without permission. Admins can also choose release channels, exclude files with an .aiexclude file, and rely on IP indemnification for code suggestions.

Copilot’s controls sit in GitHub. Enterprise owners set policies first and organisation owners work within them: which features, agents and models members can use, whether Copilot CLI is allowed, and which organisations get the cloud agent. Content exclusion, on Business and Enterprise, stops chosen files from informing suggestions, chat and code review. For a team that already manages repositories, permissions and audit in GitHub, that is one fewer console.

Gemini Code Assist vs Claude Code

Claude Code is an agent first, not an editor assistant with an agent mode. Anthropic’s overview describes it working in the terminal, IDE extensions, a desktop app and the browser, reading the codebase, editing files and running commands. It has no inline completions; you ask for outcomes and review diffs. Choose Code Assist when completions in the editor and Google Cloud administration matter most. Choose Claude Code for long, multi-step agent work, and see Claude Code vs GitHub Copilot for the same trade-off against Copilot.

Gemini Code Assist vs Codex

OpenAI’s Codex spans a CLI, an IDE extension for VS Code-compatible editors, JetBrains IDEs and Xcode, the ChatGPT desktop app and Codex cloud, which runs tasks in containers on OpenAI’s side. Like Claude Code, it is an agent rather than a completion engine, and it is bought through ChatGPT plans or used with an API key. Code Assist has no cloud agent of its own; Google’s asynchronous agent is Jules, compared in Google Jules vs Codex. For Codex against Copilot, see OpenAI Codex vs GitHub Copilot.

Which one fits your team

• You run on Google Cloud and want the assistant under the same IAM, perimeter and logging: Gemini Code Assist.
• Your code, reviews and permissions live in GitHub: Copilot, for code review and the cloud agent as much as the editor.
• Developers use Visual Studio, Eclipse or Xcode: Copilot.
• You build Android apps or work in Cloud Workstations: Code Assist fits those tools directly.
• You want suggestions shaped by your private repositories: Code Assist Enterprise’s code customisation, or Copilot’s repository and organisation instructions.

One board for whichever assistant you pick

Both agent modes take MCP servers, so both can work from the same task list. Add fenbs at https://fenbs.ai/api/mcp: in ~/.gemini/settings.json for Code Assist in VS Code, or in your editor’s MCP settings for Copilot, once the policy allows it. Sign in through the browser and tick what the assistant may do, or use a token you issue under Settings with a name, scopes and an optional expiry. The assistant then reads tasks in Next Up, moves what it starts to In Progress, sets a test status with notes, and moves finished work to Completed, and History records each change under its name. A team that has not settled on one assistant can run both against the same board.