Web Sessions

Conduit includes a built-in web session type for accessing web-based management consoles, appliances, and dashboards directly alongside your other remote connections.

Browser Engines

Conduit supports two browser engines for web sessions, letting you choose the best fit for each connection.

Chromium (Default)

The default engine uses Electron's WebContentsView (native Chromium). This provides full browser compatibility and security isolation, so even complex web applications render correctly and reliably. Chromium is available on all platforms.

Edge / WebView2 (Windows Only)

On Windows, Conduit can use Microsoft Edge's WebView2 runtime as the browser engine. This is primarily useful for accessing Microsoft 365 and Azure services on domain-joined machines, where WebView2 provides automatic single sign-on (SSO) via Windows' Web Account Manager (WAM) and Primary Refresh Token (PRT) injection.

Under the hood, a C# helper binary hosts the WebView2 control as a Win32 child window and communicates with Conduit over a named pipe. This architecture keeps the WebView2 process isolated while still allowing full navigation control, screenshot capture, and input forwarding from the main application.

Engine Selection

You can choose the engine at two levels:

• Per-entry — In the web session entry settings, select Auto, Chromium, or Edge – WebView2.
• Global default — In Settings → General, set the Web Session Engine to control the default for all new and existing entries that are set to Auto.

The resolution chain is: per-entry setting → global default → auto. When set to auto, Conduit prefers WebView2 on Windows (if the runtime is available) and uses Chromium on macOS and Linux.

Multi-Tab Browsing

Each web session supports up to 12 tabs, giving you a browser-like experience without leaving Conduit. A sub-tab bar below the main session tab shows all open tabs with favicons, titles, and loading indicators. Click any tab to switch, or click the + button to open a new tab from the session's original URL.

• Drag-to-reorder — rearrange tabs within a session by dragging them along the sub-tab bar.
• Tab state tracking — each tab independently tracks its URL, title, favicon, loading state, navigation history, and HTTPS status.
• Close tabs — close individual tabs with the X button. Closing the last tab closes the session.

Browser Navigation Toolbar

A full browser-style navigation toolbar sits above the web content, providing back, forward, stop/refresh, and home controls alongside an editable address bar. Click the address bar to type a URL and press Enter to navigate — https:// is auto-prepended if no protocol is specified. The URL is displayed without the protocol prefix for a cleaner look, and a lock/unlock icon indicates HTTPS status at a glance. The home button returns to the session's original URL.

Security

Many management consoles and network appliances use self-signed certificates. Conduit supports certificate error bypass on a per-session basis, so you can access these interfaces without browser-level certificate warnings blocking your workflow.

Content Interaction

Capture screenshots of web sessions and extract page content programmatically. Read content by CSS selector in text, HTML, or markdown format. This is especially useful for AI-assisted automation workflows where an agent needs to understand what is displayed on the page.

Download Management

When a web page triggers a file download, Conduit intercepts it and presents a toast notification showing the file name and size. From there you can choose how to handle the download:

• Open — downloads the file to a temporary location and opens it with your operating system's default application.
• Save As — opens a native save dialog so you can choose exactly where to save the file.
• Cancel — cancels the download and cleans up any partially downloaded data.

Active downloads display a real-time progress bar with percentage complete, bytes transferred, and current transfer speed, so you always know exactly where things stand. Multiple downloads can run simultaneously — each gets its own progress indicator.

Autofill

Configure CSS selector-based autofill for username, password, and submit button fields. Conduit supports multi-step login flows with URL pattern matching, so complex authentication sequences (such as a username page followed by a separate password page) can be automated.

Autofill Bar

When a web session has a linked credential, an autofill bar appears with Fill and Pick controls. The bar shows a status indicator (idle, filling, success, or error) so you always know the current state. Click Fill to populate the form using saved selectors and the entry's stored credentials.

Autofill Selector Picker

Click Pick to launch a guided 3-step wizard that walks you through selecting the username field, password field, and submit button directly on the page. Click any element to capture its CSS selector, then review and confirm your picks before saving. You can skip any step or click Done early if you only need a subset of the fields — all three are not required.

Saved selectors are persisted to the credential entry, so the next time you open the session you can fill the form with a single click.

Open in External App

From the entry dashboard, click Open in External App to open the web session URL in your system's default browser. This is handy when you need full browser features like extensions, developer tools, or bookmarks for a particular page.

MCP Integration

AI agents connected via the MCP server can interact with web sessions programmatically. Web sessions go beyond RDP and VNC parity, providing 21 tools across three interaction models including full tab management.

Coordinate-Based Interaction

These tools mirror the RDP/VNC interaction model. The agent takes a screenshot, identifies elements visually, then clicks or types at specific coordinates. Coordinates are specified in screenshot image space and automatically scaled to viewport CSS pixels.

• Take screenshots of the current page
• Click at coordinates (left, right, middle button; single or double-click)
• Type text at the currently focused element
• Send keyboard events (press, down, up) with modifier keys
• Move the mouse cursor to specific coordinates
• Drag from one point to another
• Scroll at specified coordinates
• Query the current viewport dimensions

DOM-Aware Interaction

These tools are exclusive to web sessions and use CSS selectors instead of coordinates. They are ideal for structured pages where elements can be reliably identified by selector.

• Click an element by CSS selector (uses DOM click(), works even if the element is off-screen)
• Fill an input field by CSS selector with framework-compatible event dispatch (React, Vue, Angular)
• Discover interactive elements on the page (buttons, links, inputs, selects) with text, selector, and bounds
• Execute arbitrary JavaScript in the page context and return results

Content Reading

• Read page content by CSS selector in text, HTML, or markdown format
• Navigate to any URL

Tab Management

AI agents can manage tabs within a web session, enabling multi-page workflows without opening separate connections.

• List all tabs with their URLs, titles, and active status
• Create new tabs (up to 12 per session), optionally navigating to a URL
• Close tabs and switch between them
• Navigate back, forward, and reload the active tab