GitHub Copilot · AI coding built your way (original) (raw)

From strategy to code, GitHub Copilot in Visual Studio Code keeps every project moving forward.

VS Code with TypeScript code on the left and Copilot plan steps with action buttons on the right.

Features

Put your agent to work

Partner with Copilot to plan, edit, and build together.

Strategy before code

Plan mode lets you review and approve the blueprint before the agent starts coding.

Explore plan mode

From prompt to project

Agent mode helps make changes at scale by analyzing code, proposing edits, running tests, and validating results across multiple files.

Explore agent mode

Configure once, align always

AGENTS.md enables you to share project-specific instructions and keep all agents in sync with your team’s coding practices.

Explore custom instructions

Harness your agents

Choose the right agent for the job, from Copilot to a growing ecosystem of third-party agents and platforms.

A VS Code window shows a sidebar with agent sessions. Under OpenAI Codex Agent, a button says "Start a coding session."

The right agent for every challenge. Choose from a growing ecosystem of third-party coding agents over the coming months to tackle any challenge, right from your editor.

Let Copilot work in parallel

Hand off tasks to Copilot and get a pull request back while staying in the flow.

Explore how in VS Code

Native MCP server integration

Discover and install custom MCP servers from the GitHub MCP Registry, without leaving VS Code.

Add an MCP server in VS Code

Multi-platform support

Work with Copilot wherever you code, whether natively in Visual Studio Code, or through extensions across Visual Studio, JetBrains, Eclipse, Xcode, and Neovim.

Explore extensions

Built for performance and choice

Copilot delivers instant, context-aware performance and gives you the flexibility to choose the right model for your work.

Since September 2025, Copilot delivers 2x higher throughput, 37.6% better retrieval, and an 8x smaller index size for faster, more accurate results.

Read the blog

Bar chart comparing models: Previous Model score 0.484 vs Copilot-embedding score 0.670.

Get the most out of GitHub Copilot in your IDE