Icefort Circuit

You can automate both web browsers and Electron desktop applications with the Circuit MCP server suite. It provides AI-friendly interfaces to drive web and desktop actions, capture snapshots, and manage sessions, enabling precise, end-to-end automation workflows for AI agents and developers.

How to use

Install and configure the Circuit MCP servers in your AI agent’s MCP configuration. You’ll enable web and desktop automation, then issue common actions to launch browsers or Electron apps, navigate pages, interact with elements, and capture snapshots for AI processing.

How to install

Prerequisites: ensure you have Node.js and npm installed on your development machine.

// Step 1: Create a new project folder for your MCP configuration
mkdir circuit-mcp-project
cd circuit-mcp-project

// Step 2: Initialize a new npm project
npm init -y

// Step 3: Add the MCP configuration that enables both web and Electron automation

Confi guration for dual-engine automation

{
  "mcpServers": {
    "circuit-web": {
      "command": "npx",
      "args": ["@icefort/circuit-web@latest"]
    },
    "circuit-electron": {
      "command": "npx",
      "args": ["@icefort/circuit-electron@latest"]
    }
  }
}

First commands you can run

Once you have the configuration in place, you can start automating tasks by launching a web browser or an Electron app and issuing actions to control navigation, UI interactions, and data capture.

Additional sections

This server supports rich web and desktop automation capabilities, including AI-optimized snapshots, multi-tab handling, network and console monitoring, content extraction, and test code generation from recorded actions. You manage sessions, target elements with accessibility refs, and perform complex interactions like file uploads, dropdown selections, and keyboard inputs. The system can also automate dialogs, manage windows, and provide robust error handling for reliable automation workflows.

Practical usage patterns

• Launch a web browser with AI-friendly settings, then navigate to a URL and rely on auto-snapshots with element refs to guide subsequent actions.

Security and best practices

Adopt least-privilege access for automation, isolate automation sessions, and use SIEM/logging to monitor automation pipelines. Ensure sensitive data handled by automation is treated securely and that access to automated UI actions is restricted to trusted agents.

Troubleshooting

If you encounter issues starting automation, verify your MCP configuration is loaded correctly, confirm that both web and Electron servers are reachable by your agent, and check for compatible versions of the MCP packages used by your setup.

Notes on architecture

The dual-engine design enables runtime app selection and isolated concurrent sessions, with TypeScript support, robust error handling, and optimized performance for AI-driven automation.

Tools and capabilities overview

The Circuit MCP suite exposes a comprehensive set of tools for web and desktop automation, including launching browsers and apps, navigating pages, interacting with elements, capturing snapshots, and extracting content. You can monitor network activity and console output, generate tests from actions, and manage dialogs and state waiting.

Available tools

browser_launch

Launch browser with AI optimizations and returns a session ready for actions.

browser_navigate

Navigate to a URL and automatically capture a snapshot with element references.

browser_tab_new

Create a new browser tab within a session.

browser_tab_list

List all open browser tabs with their details.

browser_tab_select

Switch to a specific tab within a session.

browser_tab_close

Close a specific tab.

click

Click an element and capture an updated snapshot with element references.

type

Type text into an element and capture a snapshot.

hover

Hover over an element and capture a snapshot.

drag

Drag an element to a target element.

key

Press a keyboard key with optional window scope.

select

Select a dropdown option by value.

upload

Upload a file to an input field.

back

Navigate backward in history.

forward

Navigate forward in history.

refresh

Reload the current page.

screenshot

Take a compressed screenshot and save to a path.

snapshot

Get the accessibility tree with element references.

pdf

Generate a PDF of the current page.

content

Get HTML content of the page.

text_content

Get visible text content from the page.

evaluate

Execute JavaScript in the page context.

wait_for_selector

Wait for an element to appear within a timeout.

close

Close the browser session.

app_launch

Launch an Electron app with AI optimizations and return a session.

get_windows

List windows with type identification for the Electron app.

ipc_invoke

Call IPC methods within the Electron app.

fs_write_file

Write a file to the local filesystem from the Electron context.

fs_read_file

Read a file from the local filesystem within the Electron context.

keyboard_press

Press a key with optional modifiers in the Electron app.

click_by_text

Click an element by its visible text.

click_by_role

Click an element by accessibility role.

click_nth

Click the nth matching element by selector.

keyboard_type

Type text with a delay to simulate natural input.

add_locator_handler

Handle modals or popups by locator and action.

wait_for_load_state

Wait for a specific page load state.

smart_click

Perform a smart click using text, refs, or CSS strategy.

browser_console_messages

Get console logs from the web context or Electron app.

browser_network_requests

Get network requests for debugging and inspection.