Inter-office memo
- To
- Everyone on GitHub Copilot
- From
- Mo, Chief Executive
- Date
- November 18, 2026, 3:14 AM CST
- Re
- Browser use in GitHub Copilot
GitHub Copilot can now use your browser. How to turn it on, and what to watch
On November 17, 2026, GitHub put browser use in public preview in Copilot CLI, the Copilot app, VS Code and JetBrains IDEs, so Copilot can test your web app, fill in forms and read pages in your signed-in Chrome or Edge. It’s off by default. You need the GitHub Copilot extension from the Chrome Web Store and Node.js 22 or later, then turn it on in your client. GitHub hasn’t named a plan it needs.
What GitHub released on November 17
On November 17, 2026, among its Microsoft Ignite releases, GitHub’s changelog said “browser use is now available in public preview” in GitHub Copilot CLI, the GitHub Copilot app, VS Code and JetBrains IDEs. Its docs put it in one line: “Copilot can interact with your browser to help you test web applications and complete browser-based tasks.”
The Copilot app’s Ignite release notes give the uses: “Ask Copilot to fill out forms, pull information from webpages, or test and debug the web app you’re building.” It “works in Chromium-based browsers, including Google Chrome and Microsoft Edge, with the GitHub Copilot extension for Chrome,” and “runs in an isolated tab group.” It isn’t a separate browser: it works in the one you already use, signed in where you are. “Browser use is off by default,” and as a preview it’s “subject to change.”
The same notes put Azure DevOps repositories in the Copilot app in public preview and made GitHub Enterprise Server accounts there generally available. GitHub also added MAI Code 2 to Copilot that day.
What you need
| Requirement | What browser use needs |
|---|---|
| Client | Copilot CLI, the Copilot app, VS Code or a JetBrains IDE |
| Browser | Google Chrome 127 or later, or Microsoft Edge with Chrome also installed |
| Extension | GitHub Copilot, from the Chrome Web Store |
| Node.js | Version 22 or later, with node on your PATH |
| In the Copilot app | Computer use turned on first |
| At work | Not blocked by your admin’s features.browserUse setting |
| Plan | Not stated by GitHub |
The extension was at version 0.2.1, updated November 17, with 1,000 users when we read its listing. It says it “Connects local GitHub Copilot agents to a Chromium browser,” and that “Browser Use requires a supported GitHub Copilot client, Node.js 22 or later, and Chrome 127 or later.” Edge works only with Chrome alongside it: the docs list “Microsoft Edge (Google Chrome must also be installed).”
How to turn it on
- Install Node.js 22 or later. All three how-to pages say the browser use plugin runs on it, so check with
node --version. - Install the GitHub Copilot extension in Chrome or Edge from the Chrome Web Store.
- Switch it on in your client, then start a new session:
| Client | Where to turn it on |
|---|---|
| Copilot CLI | Enter /plugin, find browser-use and enable it |
| Copilot app | Settings > Computer Use, turn on Enable Computer Use if it’s off, then Enable Browser Use |
| VS Code | Settings, search for “browser use,” turn on Browser Use: Enabled |
| JetBrains IDEs | Settings > GitHub Copilot > Chat, turn on Enable Browser Use |
- Check the connection. Click the extension’s icon and confirm it says Connected.
- Ask for something in the browser. GitHub’s example is “Open github.com in my browser.” Copilot picks browser use on its own for browser tasks, or you can call it with
/browser-use.
If it doesn’t connect, the CLI guide suggests checking that the plugin is enabled and the extension says Connected, then running /browser-use:doctor for guided checks. In the Copilot app on macOS 27, the extension card can wrongly read “Not installed”; GitHub says browser use still works, and the extension connects a few seconds after you send a message.
What it can reach, and how to keep it in check
The docs are plain about the risk: “Browser use can interact with websites where you are already signed in.” That’s the point of it, and the reason to be careful. “Ambiguous instructions or unexpected webpage content can cause unintended actions that affect your data or connected accounts,” and it “can select the wrong control, enter text in the wrong location,” or “have difficulty with dynamic pages and complex workflows.”
- Approve sites one at a time. “By default, GitHub Copilot requests approval before connecting to a site,” the extension’s listing says. Its settings let you “require approval, always allow, or block access for specific domains.”
- Keep sign-ins and sensitive steps to yourself. The listing says “Sign in and complete sensitive steps yourself,” and the docs say “Avoid asking Copilot to submit real credentials or make changes you have not reviewed.” Their own example asks Copilot to show a drafted email before it’s sent.
- Mind what’s on screen. “Webpages may display sensitive information, including information about other people,” so “Only use browser use with websites and tasks whose visible content you are comfortable providing as context to Copilot.” Point it at a test app where you can.
- At work, your admin decides. In managed settings, “Set the key to
falseto prevent users from enabling browser use.” The CLI, the Copilot app, VS Code and JetBrains honor it.
Browser use or computer use
They’re separate. Computer use lets Copilot click and type in desktop apps, for software with no API, command line or MCP server, and it’s in Copilot CLI and the Copilot app on macOS and Windows only. GitHub’s docs say that when a “dedicated browser tool can complete the task directly,” it gives “more structured information and predictable results,” so for websites, use browser use. In the Copilot app you need both on, since browser use there depends on computer use.
How it compares
Here’s how Copilot’s browser use sits beside the other ways to give a coding agent a browser, by each maker’s docs on November 18:
| Way | Works in | Browser | Your sign-ins | What it needs |
|---|---|---|---|---|
| Copilot browser use | Copilot CLI, Copilot app, VS Code, JetBrains | Chrome, or Edge with Chrome installed | Yes | GitHub’s extension, Node.js 22; plan not stated |
| Claude in Chrome | Claude Code’s CLI and VS Code extension | Chrome, Edge and other Chromium browsers; not WSL | Yes | A direct Anthropic plan: Pro, Max, Ultra, Team or Enterprise |
| ChatGPT’s browser | ChatGPT on the web and in its desktop app, not Codex CLI or the IDE extension | Its own, or your Chrome, Edge, Brave, Opera or Vivaldi through an extension | Only through the extension | ChatGPT; no plan listed |
| Playwright MCP | Any MCP client, Copilot CLI and VS Code included | Its own, or your running Chrome or Edge with --extension |
Only with its extension | Free and open source |
Claude in Chrome starts with claude --chrome and “shares your browser’s login state.” ChatGPT’s browser “uses a browser profile that is separate from your regular browser,” and OpenAI says “Browser isn’t available in Codex CLI or the Codex IDE extension.” Playwright MCP, Apache-2.0, works in Copilot through /mcp add in the CLI or ~/.copilot/mcp-config.json, and it “Uses Playwright’s accessibility tree, not pixel-based input.” It isn’t tied to one agent, and it doesn’t need your signed-in browser. Our guide to giving agents a browser goes through nine ways.
What GitHub hasn’t said
- Which plans get it. Neither the announcement nor the docs name one. The Copilot app “is available for all Copilot plans,” and Free “supports CLI and agent mode,” by GitHub’s plans page.
- What it costs. Nothing says browser use is billed on its own. It runs inside Copilot’s agent sessions, and the plans page describes AI credits as “Credits used when interacting with Copilot features.”
- Whether cloud sessions can use it. The extension connects “local GitHub Copilot agents,” while since Universe “every session runs Copilot in a secure cloud sandbox” by default. The managed setting’s table marks Copilot in the cloud as unsupported. GitHub hasn’t said how the two meet.
- Linux in the Copilot app. The app runs on macOS, Linux and Windows, but there browser use needs computer use, which GitHub documents for macOS and Windows. The CLI and the IDEs don’t carry that condition.
- Other browsers. Only Chrome and Edge are documented, not Brave, Arc or Vivaldi.
- What the extension sends. Its store listing declares “User activity” among the data it handles and points to GitHub’s privacy statement. GitHub hasn’t said when browser use leaves preview.
If you use other agents too
Enjoy doesn’t run Copilot’s own agents, so none of this changes inside it. It runs Claude Code, Codex, Grok Build, Antigravity, Muse Code, and local models through OpenCode, and each gets a built-in browser beside its conversation, where the agent can open your site, read the page, click, fill in forms and scroll, with nothing to set up.
9.Questions.
Can GitHub Copilot use my browser?
- Yes, since November 17, 2026, in public preview. In Copilot CLI, the Copilot app, VS Code and JetBrains IDEs, Copilot can open sites, fill in forms, read pages and test the web app you’re building, in Chrome or Edge, through GitHub’s Copilot extension. It stays off until you turn it on.
How do I turn on browser use in GitHub Copilot?
- Install Node.js 22 or later and the GitHub Copilot extension from the Chrome Web Store, then switch it on in your client. In the CLI, enter /plugin and enable browser-use; in the Copilot app, turn on Settings > Computer Use > Enable Browser Use; in VS Code, the Browser Use: Enabled setting; in JetBrains, Settings > GitHub Copilot > Chat > Enable Browser Use. Then start a new session.
Which Copilot plans get browser use?
- GitHub hasn’t said. Its announcement and docs name no plan. The Copilot app “is available for all Copilot plans,” and GitHub’s plans page says Free “supports CLI and agent mode,” with limited chat and agent usage, while paid plans draw agent work from monthly AI credits.
Does Copilot’s browser use work in Edge or Brave?
- Edge, yes, but GitHub’s docs add “Google Chrome must also be installed.” The docs name only Chrome and Edge, so Brave, Arc and other Chromium browsers aren’t documented. The extension needs Chrome 127 or later.
Is it safe to let Copilot use my signed-in browser?
- It can act on any site where you’re signed in, so be deliberate. By default the extension asks before it connects to a site, you can block or always allow a domain, and the extension’s own listing says “Sign in and complete sensitive steps yourself.” Try it on a test app first, and review what it did.
Why does the Copilot app say the extension isn’t installed?
- On macOS 27 that’s a known bug. GitHub’s docs say browser use still works there. Start a session and send a message, and the extension connects within a few seconds. Elsewhere, check that the extension shows Connected, and in the CLI run /browser-use:doctor.