# Ui.Vision MCP [Health: Active]

**Category:** 📂 Browser Automation  
**Repository:** https://github.com/A9T9/RPA  
**GitHub Stars:** 2103  
**npm Downloads (last month):** 756  
**Views:** 5  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/ui-vision-mcp

## Description
MCP server for browser and desktop automation: OCR, image recognition, real mouse and keyboard input

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "ui-vision-mcp": {
    "command": "npx",
    "args": ["-y","uivision-mcp-bridge"]
  }
}
```

## Documentation

## What Ui.Vision MCP MCP server does

Ui.Vision MCP connects an MCP client to the Ui.Vision browser extension through the `uivision-mcp-bridge` package. The bridge allows an assistant to inspect a page, author or modify a macro, execute the macro, and examine its output through logs and screenshots. Automation runs inside Ui.Vision in the connected browser.

The underlying extension supports browser workflows such as filling forms, extracting web data, downloading reports, and repeating recorded tasks. It also supports website testing with recorded interactions, replay, Selenium IDE commands, and different input data. OCR can identify text on screen, while image recognition can locate controls when selectors are not practical.

Desktop automation is available through Ui.Vision's additional XModules components. Those components support visual interaction with desktop applications and remote desktop interfaces, including mouse and keyboard input.

## How it works

The Ui.Vision MCP MCP server acts as a bridge rather than a standalone browser driver. An MCP client sends an instruction to Ui.Vision, which performs the operation in the browser extension. A typical request can open a page, read its heading, and save the sequence as a reusable macro.

Macros can be recorded with the extension, authored through the command table, or generated with JavaScript using the `uiv.*` API. They execute sequentially, so the API is used without `async` or `await`. The API covers page elements, visual matching, OCR, browser and desktop input, screenshots, tabs, CSV files, and downloads. Classic command-table operations can also be represented with `uiv.run('command', 'target', 'value')`.

After execution, the assistant can use the available logs and screenshots to assess the result and revise the macro. Saved macros can then be run again outside the initial interaction.

## Setup and configuration

Install the Ui.Vision extension from the Chrome Web Store, Microsoft Edge Add-ons, or Firefox Add-ons. Building the repository is not required for normal use. Install XModules separately if the workflow needs desktop input or other native capabilities.

Node.js and npm are required for the bridge installer. Run:

```bash
npx uivision-mcp-bridge --setup
```

The installer registers Ui.Vision with supported MCP clients it finds on the machine. Next, enable the MCP bridge under **Ui.Vision → Settings → AI**, follow the connection instructions in the setup guide, and restart the MCP client. The repository's `mcp/` directory contains the bridge source, configuration information, and troubleshooting material.

## Tools and capabilities

The Ui.Vision MCP MCP server exposes capabilities for:

- Inspecting a page through the connected Ui.Vision session.
- Creating and editing reusable browser macros.
- Running macros and checking execution logs.
- Reviewing screenshots produced by automation.
- Using OCR to find visible text.
- Locating controls with image recognition.
- Automating browser and, with XModules, desktop interactions.
- Working with screenshots, tabs, CSV files, and downloads through the JavaScript API.

## Limitations and notes

The bridge depends on the Ui.Vision browser extension and a configured MCP client; it does not replace the extension. Desktop workflows need XModules in addition to the extension. The installer requires Node.js and npm. Macro execution follows Ui.Vision's sequential runtime conventions, so generated JavaScript should use the supported `uiv.*` API and avoid `async` or `await`.

The repository is primarily the extension source and bridge implementation. For ordinary installation, use the published browser extension rather than building the repository. The README identifies the extension as open source and free for personal and commercial use, but it does not specify a short license identifier.

_Full upstream README: https://allmcps.com/mcp/ui-vision-mcp/readme_