The open platform for AI agents
Describe a task and let agents plan it and keep going autonomously until it's done. Your choice of agents and models, regardless of where you work.
Download
From prompt to finished task
Give an agent a goal and it does the work — researching, editing, running commands, and checking its own results.
Plan before you build
Let the agent research and plan the work first. Review and approve the strategy before it writes a single line of code.
Run autonomously
Agents make edits across files, run terminal commands, and iterate on failures until the task is complete, with you in control.
Work in parallel
Don't wait on one agent to finish. Kick off multiple tasks at once, across projects, and keep everything moving forward together.
Agents work where you do
When you use Copilot, your agents follow you across VS Code, the command line, the browser, and your phone. Start a task anywhere, finish it anywhere.
Work however the task demands
Pick your agent, GitHub Copilot or a third-party agent running locally or in the cloud, then move freely between an agent-first flow for high-level tasks and a code-first flow for the details, all in one session.
And whenever you want to edit code yourself, the complete VS Code editor is right there, with full IntelliSense, refactoring, and debugging included.
Move from issue to commit without leaving your shell
Run a GitHub-native agent right in your terminal. It works directly with your issues and pull requests, plans the change, edits across files, and runs commands, all without breaking your command-line flow.
When a task needs a closer look, run /ide to open it in VS Code and refine the
code directly, then jump back to the terminal whenever you like.
Manage agent work from the browser
Stay on top of your agents from any browser, no setup required. On github.com, assign issues to Copilot, review agent pull requests, and track progress, then pull a branch into VS Code when you want to take over.
Need your own environment? vscode.dev opens a secure tunnel to your development machine, letting you track and manage your running agent sessions in the browser with your local code and tools within reach.
One workspace for your agent work
Launch work from GitHub, track progress across multiple agents, review changes, and merge completed work, all from one desktop workspace built natively on GitHub.
Explore the Copilot AppChoose the right agent for the job
Use GitHub Copilot, your own custom agents, or the third-party agents you already rely on. Run them locally on your machine or in the cloud for scale.
GitHub Copilot
Run Copilot agents locally in VS Code with full access to your codebase, or delegate to the cloud and let them work in the background and open a pull request for your team.
Best for: everyday development — from interactive, in-the-loop tasks to async work you hand off and review later.
Custom agents
Create custom agent personas with their own instructions, tools, and skills — tuned to your team's conventions and the tasks you do most.
Best for: repeatable workflows where a purpose-built agent should follow specific rules and use specific tools.
Third-party agents
Run agents from Anthropic, OpenAI, and others locally or in the cloud — right alongside the tools you already know.
Best for: teams with existing agent preferences who want VS Code's editing power.
All your agents, one view
Monitor and manage all your agent sessions from a unified view, regardless of where they run.
Use the Agents window for an agent-first workflow, where you instruct agents with high-level tasks.
Use the Chat view for a code-first workflow, where agents run alongside the code editor to help you accomplish coding tasks. Switch freely between the two at any time.
Manage agent sessions
Swap the model, keep the workflow
Choose from dozens of models across OpenAI, Anthropic, Google, and more.
Bring your own API key to use any model from any provider, including models you host yourself.
Learn more about language models
Native MCP integration
Model Context Protocol support is built in. Connect agents to your databases, APIs, and services through MCP servers — install them from the marketplace or bring your own, and manage them right inside the editor. No glue code, no extra tooling.
Explore MCP servers
Agents that work your way
Define custom instructions, add skills, and build custom agents so they follow your team's conventions and workflows. Browse the marketplace to install ready-made customizations from the community.
Learn how to customize your agents
Real tasks. Real results.
Agents help everyone on the team, from prototyping ideas to shipping features as pull requests.
Build a feature end-to-end
Describe a feature in natural language. The agent scaffolds, implements across files, and runs tests to verify.
Debug and fix failing tests
Point an agent at a failing test. It traces the root cause, applies a fix, and re-runs to confirm.
Refactor or migrate a codebase
Plan a migration between frameworks or APIs. The agent applies coordinated changes and verifies with builds.
Ship prototypes, not specs
Product managers and designers turn user feedback, a Figma file, or a rough idea into a working pull request the team can review and iterate on.
Audit accessibility and UX
An agent launches your app with Playwright, navigates to the feature, and verifies behavior, layout, and accessibility with screenshots.
Write docs, release notes, and summaries
Summarize the day's commits across repos into API docs, changelogs, release notes, or social posts.