View Control MCP Server

What is View Control MCP Server?

View Control MCP Server is a Model Context Protocol (MCP) server that enables AI assistants to control a local PC through mouse operations, keyboard input, and screen capture. It implements the MCP protocol (JSON-RPC 2.0 over stdio) and integrates with MCP clients like Claude Desktop and Cursor.

How to use View Control MCP Server?

Install the package via npm (npm install view-control-mcp-server), then start the server with npx view-control-mcp-server. Configure a client (e.g., Cursor or Claude Desktop) with transport type stdio and command npx view-control-mcp-server. The server will then accept JSON-RPC requests on standard input/output.

Key features of View Control MCP Server

• Mouse control: get position, move, click (left/right/middle), drag‑and‑drop.
• Keyboard input: type text, press keys, and combine with modifier keys.
• Screen capture: capture the full screen and return Base64‑encoded image data.
• JSON‑RPC 2.0 protocol with stdio transport.
• Compatible with multiple MCP clients (Claude Desktop, Cursor).

Use cases of View Control MCP Server

• Automate local desktop tasks via an AI assistant.
• Test and debug UI interactions programmatically.
• Enable an AI chat tool to perform mouse or keyboard actions on the user’s machine.
• Capture screen content for analysis or reporting.

FAQ from View Control MCP Server

What is the MCP protocol used by this server?

The server uses JSON‑RPC 2.0 over standard input/output (stdio). It supports protocol version 2024-11-05 and communicates with MCP clients via a standardized set of methods.

How do I connect this server with Cursor?

In Cursor’s MCP settings, add a new server with name "View Control MCP Server", transport stdio, and command npx view-control-mcp-server. Leave the Args field empty.

How do I connect this server with Claude Desktop?

In Claude Desktop’s MCP settings, add a server with the same configuration: name "View Control MCP Server", transport stdio, command npx view-control-mcp-server.

What methods are available on the server?

The server exposes methods under mouse (e.g., mouse.move, mouse.click), keyboard (e.g., keyboard.type, keyboard.press), and screen (e.g., screen.capture), plus standard MCP methods like initialize, ping, tools/list, and tools/call.

Are there any security considerations?

Yes. Because the server controls the local PC, use it only in trusted networks and be mindful of privacy when using screen capture features.