Skip to main content
Major can use a web browser to get real work done: read websites, fill in forms, click through pages, and sign in. The browser runs in the cloud, not on your computer, and there’s no connector to set up. The browser is available to: Use the browser for sites with no API, one-off lookups, or when you ask for it.

​
Example requests

  • “Find the pricing page for Acme and summarize their plans.”
  • “Look up our company on the state business registry and tell me when the annual report is due.”
  • “Sign in to our supplier portal and download last month’s invoice.”
  • “Draft a LinkedIn message to Jane Doe asking for a call next week.”

​
What it can do

The agent drives the browser one action at a time. It can:
  • Open pages in tabs and move between them.
  • Read a page’s text, or its structure to find buttons, links, and fields.
  • Click, type, select options, and press keys.
  • Take screenshots of a page or a single element.
  • Inspect a page’s network requests and console errors.
  • Sign in with a saved login.
The agent can’t see a live view of the browser, and you don’t get a link to one either. Before it sends, posts, buys, submits a form, deletes something, or changes account settings, the agent confirms the exact recipient and content with you, unless you already gave both. Read-only browsing, such as searching or reading, needs no confirmation.

​
What it can’t do

  • Open Major apps. The browser isn’t signed in to Major, so a deployed app only shows its sign-in page. Session preview URLs are blocked. To check an app, the Platform Agent uses the app’s preview and logs instead.
  • Handle every sign-in check on its own. For one-time codes or “approve on your phone” prompts, the agent asks you in the chat.
  • Go where your browser controls don’t allow. Blocked sites, allowlists, and rules are checked before every action.

​
Turning it on for an agent

Browser access is off for new agents. Turn it on in the agent editor’s Details panel under Browser access, or set browserSettings in agent.jsonc:
The same browserSettings object holds the agent’s own browser controls.

​
Browser sessions and sign-ins

A browser starts on the first browser action in a chat. Cookies and site sign-ins are kept in a browser profile:
  • Your chats share one profile. A site you signed in to in one chat stays signed in in your other chats.
  • An agent’s runs share one profile for that agent, whoever runs it. Your own profile is never used for an agent’s runs.

​
Saved logins

When the browser reaches a sign-in page and has no login saved for that site, the agent shows a login card in the chat (or in Slack). The card opens a separate Major page where you enter the fields the site asks for, such as email, username, password, or company ID. Major fills them into the sign-in form. The AI never sees the values, and the agent never asks you to type a password into the chat. On the login page:
  • Save this login is on by default, so Major can sign in to that site again later. Turn it off to use the login just once.
  • Decline tells the agent you won’t provide the login.
Also good to know:
  • Logins are saved per exact site, so www.example.com and example.com need separate logins.
  • On signup forms, Major generates a password and saves the new login for you.

​
Managing your saved logins

Open Account settings > Browser. Your saved logins are listed under Saved logins. You can add one ahead of time, add a label, edit fields, reveal saved values, or delete a login. Deleting a login doesn’t sign the browser out of sites it’s already signed in to. Only your own chats can use your saved logins. Organization admins can’t see or use them, and agents never use them.

​
An agent’s saved logins

An agent with browser access has its own saved logins. They belong to the agent, not to a person, and only that agent’s runs fill them in. When you build an agent with the Platform Agent or an MCP client, it can send you a link to save a login to the agent ahead of time. The page shows which site the login is for, and you need edit access to the agent to save it.
Anyone who can run the agent can have it sign in as that account. Only save logins to an agent when that access is fine to share with everyone who can run it.

​
Who can use the browser

The browser is available to organization Members, Builders, and Admins. App viewers don’t get it. In a custom role, it’s the Use the hosted browser capability. The browser only runs in Major chats and agent runs. An MCP client like Claude Code or Cursor doesn’t get the browser tool.

​
Billing

Browser time uses your organization’s credits. You can see agent browser spend on the agents dashboard under Analytics.

​
Next steps

Browser Controls

Choose which sites the browser can visit and what it can do there.

Agents

Build an agent and give it browser access.