GitHub Copilot Vision generally available announcement Image: GitHub / github.blog
by VibecodedThis

GitHub Copilot Vision Is Now GA, and So Are Browser Tools for Agents

GitHub made two GA announcements on July 1: Copilot Vision now works on all plans with no admin action required, and browser tools let agents drive live web apps directly from VS Code.

Share

GitHub shipped two general availability announcements on July 1 that had both been in preview for a while: Copilot Vision and browser tools for agents in VS Code.

Copilot Vision

Copilot Vision lets you attach images and PDFs to chat prompts so Copilot can reason about them alongside your code. The feature was previously gated behind an Editor Preview Features policy that admins on Business and Enterprise plans had to enable at the org or enterprise level.

That gate is gone. Vision is now on by default for every Copilot subscriber: Free, Pro, Pro+, Business, and Enterprise. No policy changes, no admin action required.

It works across VS Code, github.com, and the CLI. GitHub retains image and PDF attachments for roughly 24 hours on Business and Enterprise to provide the service.

The practical use cases are the obvious ones: paste in a screenshot of a failing UI, drop in a PDF spec, share a diagram of an architecture you’re trying to implement. The model sees what you see.

Browser Tools for Agents

Also GA on July 1: browser tools for GitHub Copilot agents in VS Code. Agents can now drive a real browser, navigate live web apps, and feed results back into chat. Browser tools are on by default.

What agents can do in the browser:

  • Open and navigate pages, click, type, hover, drag, and handle dialogs
  • Read page content, capture console errors, take screenshots
  • Run scripted multi-step flows

Your tabs stay private by default. An agent can’t read or interact with a page you opened unless you explicitly select “Share with Agent.” You can revoke that access at any time. Tabs the agent opens itself run in fresh sessions isolated from your cookies and everyday browsing. Agents running in parallel in the Agents window each keep their browser tabs private from one another.

Sensitive permissions like camera, microphone, location, notifications, and clipboard reads are never granted automatically. They each require your explicit approval per site. The agent can’t approve them on your behalf.

For enterprise admins, there’s a dedicated toggle (workbench.browser.enableChatTools) and the existing network domain filtering (chat.agent.allowedNetworkDomains, chat.agent.deniedNetworkDomains) applies to the browser as well.

The browser tools are available in both the standard editor window and the Agents window. Update VS Code to get them.

Sources: Copilot Vision GA, Browser Tools GA

Share