How to Let Your Coding Agent Use Your Logged-In Chrome (2026)
By Steven Van ·
A comparison of how Claude Code, Codex, Gemini CLI and Cursor can use your signed-in Chrome, with setup steps and limits for each tool.
A new guide from Browser Use compares ways to give a coding agent access to your signed-in Chrome. It covers Claude Code, Codex, Gemini CLI and Cursor, and the tools were checked against their own docs on October 25, 2026. Browser Use also makes Browser Harness, one of the options, and says so in the guide.
Which option fits which agent
- Claude Code: Claude in Chrome (claude --chrome) on a Pro, Max, Team or Enterprise plan. On an API key, Bedrock or another provider it stays off. Browser Harness, Playwright MCP's extension mode or Chrome DevTools MCP work there instead.
- Codex: OpenAI's browser extension in the ChatGPT desktop app. In Codex CLI, Browser Harness or one of the two MCP servers.
- Gemini CLI: its built-in browser agent with sessionMode set to "existing", or any MCP option.
- Cursor: its built-in Browser keeps its own cookies per workspace, not your Chrome logins. For your own Chrome, add Browser Harness, Playwright MCP or Chrome DevTools MCP to .cursor/mcp.json.
- Any MCP client: Playwright MCP in extension mode gives the agent tabs you pick. Chrome DevTools MCP in auto-connect mode, or Browser Harness, gives it your whole running Chrome.
Why not copy cookies into a fresh browser
Exporting cookies, for example with Playwright MCP's --storage-state, works on some sites. The guide says it can fail where sessions are tied to your device. With Device Bound Session Credentials, Chrome holds a private key in secure hardware such as a TPM, so copied cookies expire and can't be renewed on another machine. A fresh profile also starts logged out, so every login and second factor has to be repeated. The Chrome DevTools MCP docs note that some accounts may block sign-in when the browser is controlled through WebDriver.
The older shortcut is closed too. From Chrome 136, --remote-debugging-port and --remote-debugging-pipe are ignored on the default Chrome data directory and need --user-data-dir pointing elsewhere.
The options in detail
Claude in Chrome works with Claude Code only. Install the extension (version 1.0.36 or later), sign in with /login, then run claude --chrome. Claude opens new tabs in a tab group and shares your login state. It pauses at login pages and CAPTCHAs, and it stays off in WSL.
Playwright MCP in extension mode uses the --extension flag to connect to your running Chrome or Edge through the Playwright Extension. It works with any MCP client. A page opens the first time so you can pick the tab. Each connection needs approval unless you set PLAYWRIGHT_MCP_EXTENSION_TOKEN in the server's env. It is free and Apache-2.0.
Chrome DevTools MCP attaches to your running Chrome with --autoConnect on Chrome 144 or later, after you turn on remote debugging at chrome://inspect/#remote-debugging. Chrome asks you to allow the session and shows an automation banner. It connects to the default profile and sees every open window in it. Usage statistics are collected by default, and --no-usage-statistics turns that off.
Gemini CLI's browser agent is off by default. It needs remote debugging enabled (Chrome 144 or later) and sessionMode set to "existing" in settings.json. The default, "persistent", launches Chrome with its own profile. Form filling always asks for confirmation, allowedDomains limits where it can go, and maxActionsPerTask stops it after 100 tool calls by default.
Codex in the ChatGPT desktop app uses OpenAI's browser extension with Chrome, Edge, Brave, Opera or Vivaldi. Install it from Settings, Computer Use. The docs only describe the desktop app, and Codex CLI isn't mentioned.