Browser Use in the Desktop App

The Browser view lets you visit websites from a task detail page in LatchLoop Desktop. You can browse yourself or explicitly ask an eligible local agent to help with a web workflow.


Browser tabs in task detail

To browse from a task:

  1. Open the task in LatchLoop Desktop.
  2. Open the Browser view from the task detail page.
  3. Enter a URL or use an existing tab.
  4. Open more tabs as needed, switch among them, navigate normally, or close them when you are done.

Your browser tabs are task-specific. LatchLoop saves the tab list, active tab, titles, and URLs for that task and restores them on the same desktop installation. Opening another task shows that task’s own set of tabs.

The Browser view is different from Preview. Use Browser for arbitrary websites and web workflows. Use Preview to run and inspect a local app from a mapped repository.


How tabs and sign-ins are saved

Tabs and website sessions have different scopes:

  • Tabs belong to a task. Each task keeps its own tab list and active tab.
  • Cookies and signed-in sessions are normally shared across tasks in the same project. After you sign in to a site in one task, another task in that project can usually use the same signed-in session.
  • Unrelated projects stay separate. A login saved for one project’s browser is not normally available to another project.

LatchLoop does not automatically copy passwords, cookies, or login state from Chrome or another browser. Sign in manually inside LatchLoop’s Browser view.


Let a local agent interact with a website

When an eligible agent run is in local mode, you can explicitly ask it to work with the Browser tabs for the current task. For example, ask it to open a page, complete an ordinary form, or verify that a change appears on a site.

At your request, the local agent can:

  • inspect the visible page
  • navigate to another page
  • switch among the task’s existing tabs
  • click controls and fill ordinary form fields
  • scroll through the page
  • verify the current page state

This also works on sites where you have already established a login session in the project browser. The agent does not control the browser constantly or in the background: request the web workflow when you want it performed. While automation is active, the Browser UI indicates that the agent is controlling a tab. Select Stop automation whenever you want immediate control back.


Cloud mode versus local mode

Cloud and local agents may both use their available web search or page-fetching capabilities to access public web content. Those capabilities are separate from the tabs open in LatchLoop Desktop.

CapabilityCloud modeLocal mode
Access public web content with available web toolsYesYes
List, inspect, or read your open desktop Browser tabsNoYes, when requested and available for the run
Control a task’s desktop Browser tabsNoYes, when requested and available for the run
Use the authenticated session stored on your computerNoYes, through the task’s Browser tabs

Web search or page-fetching tools cannot see an open tab or access its authenticated contents. If a workflow depends on a tab or signed-in session in LatchLoop Desktop, run the task in local mode and ask the agent to use the Browser.


Security and user control

  • Enter credentials and complete sign-in yourself. Do not ask the agent to enter or extract passwords, cookies, API keys, MFA codes, payment details, or other secrets.
  • The agent may stop and return control when it reaches an authentication wall, payment, destructive action, or sensitive account page.
  • Review important actions before they are submitted or finalized.
  • Select Stop automation whenever you want to end browser automation immediately.
  • Because cookies persist within a project, sign out inside the Browser view when you no longer want that project browser to retain access to a site.

Troubleshooting

The agent says it cannot access the open tab

  • Confirm the task is running in local mode.
  • Confirm LatchLoop Desktop is online.
  • Ask the agent explicitly to use the task’s Browser tab.

A site asks for authentication

  • Sign in manually in the Browser view.
  • After sign-in is complete, ask the agent to continue.

A login is not available in another task

  • Confirm both tasks belong to the same project.
  • Remember that sign-in sessions can be shared within a project, but tab lists remain task-specific.

You no longer want browser automation

  • Select Stop automation to take immediate control back.