Browser

When a task needs the web — reading documentation, checking a page, filling in a form in a local app — the engine drives a real Chromium browser. Rinari Agent shows that exact page, so you see what the agent sees.

Where it is

The browser lives in the session dock, next to Files and Terminal, both in Normal view and in each board pane. When the engine opens a browser for a session you’re looking at, the Browser tab can reveal itself; otherwise the tab shows an indicator.

The panel embeds the native Chromium target the engine is actually using — the same page, storage and lifecycle — not a copy. If a session has more than one page, choose which one to see.

Take control

The panel shows who is driving: Agent is driving or You are driving.

• Take control lets you use the page yourself. It pauses the agent’s changes on that page.
• Return to the agent hands it back.
• Open outside opens the page outside the app.

Lifetime

The browser survives individual turns and closes with its session or with the engine. Looking at the panel never launches or navigates a browser by itself. If the browser disconnects, its pages are gone, and reopening starts a new one; the panel says so instead of showing a stale page.

Permissions

Reading the internet is free in every mode. Interacting with a web page and uploading a file from the browser are actions that can ask for approval, depending on the chat’s permissions — see Approvals and permissions. After Rinari reads external content in a turn, sending data out asks even with Full access.

Local HTML files

Previewing a local HTML file — for example, a small game you built — is a separate feature with its own viewer; it doesn’t use the engine’s browser.