What is a headless browser, and when does a real one win?

What is a headless browser?

A headless browser is a web browser that runs with no visible window: a program starts it, tells it which address to load, and reads the page back as text, a screenshot or a file. It is the same kind of software you browse with, minus the screen and the buttons.

The word "headless" means no head, no window. Tools such as Puppeteer, Playwright and Selenium start and drive one from code. Developers use them for tests, for screenshots and for reading many pages in a row, usually on a server with nobody watching. This definition is general knowledge about browsers, not a Minda fact; what the page says about Minda is checked against the app.

How does a headless browser work?

Four steps, always the same, whatever tool drives it.

• Code starts it. A script launches the browser with its window switched off and, usually, a fresh profile: no history, no cookies, no sign-ins, no saved passwords.
• Code gives the orders. Open this address, wait for that element, click it, type here, take a screenshot.
• Code reads the result. The page’s text, the number of elements that match, a picture, a downloaded file.
• It ends when the script ends. The profile goes with it, unless the script saves one.

An example: testing a checkout behind a login for a Boston shop

Priya Raman builds the web shop of a Boston shoe store. Its staging site, staging.shoes.example, needs a customer login to reach the checkout, and she wants to know that the checkout shows an error when the zip code is missing. The shop and the people in this example are made up.

1. For the nightly test she uses a headless browser. Her script starts a fresh browser, opens the login page, types the test account’s email and password, waits for the account page, opens the cart, presses Place order and reads the error. It runs the same way every night on a build server, which is what a test suite needs.
2. This afternoon she wants only a quick look at a bug a colleague saw in a real session, with a cart already full. A script would have to rebuild that state first.
3. She signs in once in Minda, in a Staging workspace, and keeps that workspace active, because the agent’s tabs open in the active one. Then she tells Claude Code: "Use the minda tools. Open the cart on our staging site, snapshot it, click Place order, take a screenshot and tell me the error on the page."
4. Minda asks for her OK before the first click. The agent opens the page in her signed-in session, in a tab she does not have to look at, clicks, takes the screenshot and reports: "Zip code is required."
5. The script stays in the build and the real browser handled the one-off look. Neither replaced the other.

What can you use a headless browser for?

• Automated tests. The same clicks through a sign-up or a checkout, every night, on a server.
• Screenshots and PDFs. A picture of a public page at a set size, taken the same way every time.
• Reading many pages. A script that loads thousands of pages and keeps one field from each.
• Pages that build themselves in the browser. The text is there only after the page’s own scripts run, so a plain download of it is empty; any browser, headless or not, runs them.
• Build servers and other places with no screen. A headless browser needs none.

How to use a real browser over MCP in Minda

1. In Minda, open Settings → Developer and turn on Let coding agents control Minda. It is off until you do, and it listens on this computer only.
2. Under Connect an agent, pick your editor or coding agent (Claude Code, Codex, Cursor, VS Code, Antigravity and more) and copy its command or config. It needs Node.js 18 or later.
3. Paste it into the tool’s MCP settings, once. The tool then starts Minda’s MCP server by itself.
4. Ask the coding agent for a first task: "Use the minda tools: open this page and take a screenshot."
5. Allow each click and typed word when Minda asks, or turn on Always allow its own tabs. The agent then acts without asking in the tabs it opened, while your own tabs, your history and a press that pays still ask.

Where does a headless browser fall short?

Four places, and each is where a real browser, the one you already use, has the advantage.

• Sign-ins. A fresh profile has none of yours, so the script has to log in itself, and a login with a code or a second step is hard to script. A real browser you are signed in to starts on the page you see, in the workspace you pick.
• Bot checks. Some sites show a check to a browser that a program drives, and a page behind a check shows no content. A real browser may meet fewer of them, but it can still meet one: a CAPTCHA is yours to pass, and Minda does not go around it.
• Extensions and settings. A fresh profile has none of your extensions, saved passwords or site settings, so a page can behave differently from the one you use.
• Seeing it. Nobody watches a headless run, so a failure is a log line or a screenshot afterwards. In Minda the agent’s tabs open in your window, and each click can ask first.

Headless browser or a real one: which for which task?

Neither is better in general. The table says which fits which job, including where the real browser in Minda does not.