MCPgee
MCPgee
MCPgeeMCPgee

MAI UI

v1.0.0•Coding Agents•stable

MAI-UI: Real-World Centric Foundation GUI Agents ranging from 2B to 235B

device-cloud-collaborationgui-agentgui-groundinggui-navigationmcp
Share:
1,809
Stars
0
Downloads
0
Weekly
0/5
View on GitHub

What is MAI UI?

MAI UI is a Model Context Protocol (MCP) server that allows AI assistants like Claude, Cursor, and VS Code to mai-ui: real-world centric foundation gui agents ranging from 2b to 235b

MAI-UI: Real-World Centric Foundation GUI Agents ranging from 2B to 235B

This server falls under the Coding Agents category on MCPgee, the world's largest MCP server directory with 33,000+ servers.

Features

  • MAI-UI: Real-World Centric Foundation GUI Agents ranging fro

Use Cases

Build GUI agents with real-world centric foundation models.
Enable device-cloud collaboration and GUI navigation automation.
Tongyi-MAI

Maintainer

LicenseApache-2.0
Languagejupyter notebook
Versionv1.0.0
UpdatedMay 22, 2026
Statushealthy
Maintenanceactive

Works with

ClaudeOpenAIwindowsmacoslinux
View SourceBrowse All Servers

Installation

Manual Installation

npx mai-ui

Configuration

Configuration Details

Config File

claude_desktop_config.json

Performance

Response Metrics

Response Time< 200ms
ThroughputMedium

Resource Usage

Memory UsageLow
CPU UsageLow

How to Set Up and Use MAI UI

MAI-UI is an open-source family of foundation GUI agent models (2B to 235B parameters) from Tongyi/Alibaba that enables real-world device interaction through natural language. The models are trained for GUI grounding and navigation tasks — locating UI elements on screen, executing cross-app workflows, and dynamically offloading complex subtasks between device and cloud. Developers integrate MAI-UI with MCP to build autonomous agents that can operate mobile or desktop interfaces without hand-crafted automation scripts.

Prerequisites

  • Python 3.9+ with pip installed
  • vLLM 0.11.0 installed exactly (other versions are incompatible)
  • transformers >= 4.57.0
  • GPU with sufficient VRAM for the chosen model size (MAI-UI-2B requires ~6 GB, MAI-UI-8B requires ~20 GB)
  • An MCP-compatible client such as Claude Desktop or a custom MCP host
1

Clone the repository

Clone the MAI-UI repository to your local machine to access model configs, requirements, and example notebooks.

git clone https://github.com/Tongyi-MAI/MAI-UI.git
cd MAI-UI
2

Install exact dependencies

Install vLLM at the pinned version 0.11.0 (critical — other versions break compatibility) and then install the remaining requirements.

pip install vllm==0.11.0
pip install -r requirements.txt
3

Download a MAI-UI model

MAI-UI-2B and MAI-UI-8B weights are publicly available on HuggingFace. Download the model you want to host locally.

huggingface-cli download Tongyi-MAI/MAI-UI-8B --local-dir ./models/MAI-UI-8B
4

Start the vLLM API server

Launch the OpenAI-compatible API server that hosts the MAI-UI model. The server listens on port 8000 by default.

python -m vllm.entrypoints.openai.api_server \
  --model ./models/MAI-UI-8B \
  --served-model-name MAI-UI-8B \
  --host 0.0.0.0 \
  --port 8000 \
  --tensor-parallel-size 1 \
  --trust-remote-code
5

Configure the MCP client

Point your MCP client at the locally running vLLM server. The server exposes an OpenAI-compatible endpoint at http://localhost:8000/v1.

{
  "mcpServers": {
    "mai-ui": {
      "command": "npx",
      "args": ["mai-ui"],
      "env": {
        "MAI_UI_BASE_URL": "http://localhost:8000/v1",
        "MAI_UI_MODEL": "MAI-UI-8B"
      }
    }
  }
}
6

Test GUI agent capabilities

Send a GUI grounding request through your MCP client to verify the agent can locate and interact with UI elements on screen.

MAI UI Examples

Client configuration

MCP client config pointing at a locally hosted MAI-UI-8B vLLM server.

{
  "mcpServers": {
    "mai-ui": {
      "command": "npx",
      "args": ["mai-ui"],
      "env": {
        "MAI_UI_BASE_URL": "http://localhost:8000/v1",
        "MAI_UI_MODEL": "MAI-UI-8B",
        "MAI_UI_MAX_TOKENS": "2048",
        "MAI_UI_TEMPERATURE": "0.0"
      }
    }
  }
}

Prompts to try

Example tasks that exercise MAI-UI's GUI grounding, navigation, and device-cloud collaboration capabilities.

- "Open the Settings app and navigate to Wi-Fi settings"
- "Find the search bar on this screen and type 'hello world'"
- "Locate the Submit button and click it"
- "Navigate to the contacts list and call John Doe"
- "Switch between the Maps and Calendar apps to check my afternoon schedule"

Troubleshooting MAI UI

vLLM fails to start or produces CUDA errors

Confirm you have exactly vLLM 0.11.0 (`pip show vllm`) and transformers >= 4.57.0. Other vLLM versions are explicitly unsupported by MAI-UI.

Model produces incorrect bounding boxes or fails to locate UI elements

Ensure temperature is set to 0.0 as specified in the runtime config. Higher temperatures destabilize GUI grounding predictions.

Out-of-memory error when loading the model

Switch to a smaller model variant (MAI-UI-2B instead of 8B) or increase tensor-parallel-size if you have multiple GPUs available.

Frequently Asked Questions about MAI UI

What is MAI UI?

MAI UI is a Model Context Protocol (MCP) server that mai-ui: real-world centric foundation gui agents ranging from 2b to 235b It connects AI assistants to external tools and data sources through a standardized interface.

How do I install MAI UI?

Follow the installation instructions on the MAI UI GitHub repository. Clone the repo, install dependencies, and add the server config to your AI client.

Which AI clients work with MAI UI?

MAI UI works with all major MCP-compatible AI clients including Claude Desktop, Claude Code, Cursor, VS Code (GitHub Copilot), Windsurf, and Cline.

Is MAI UI free to use?

Yes, MAI UI is open source and available under the Apache-2.0 license. You can use it freely in both personal and commercial projects.

Learn More About MCP Servers

Getting Started with MCP

Set up your first MCP server in minutes

MCP Setup Guide

Configure MCP in Claude, Cursor & VS Code

All MCP Tutorials

18+ hands-on guides for developers

MCP FAQ

40+ answers about Model Context Protocol

MAI UI Alternatives — Similar Coding Agents Servers

Looking for alternatives to MAI UI? Here are other popular coding agents servers you can use with Claude, Cursor, and VS Code.

Dify

★ 142.2k

Production-ready platform for agentic workflow development.

Ruflo

★ 54.0k

🌊 The leading agent orchestration platform for Claude. Deploy intelligent multi-agent swarms, coordinate autonomous workflows, and build conversational AI systems. Features enterprise-grade architecture, self-learning swarm intelligence, RAG integrat

Goose

★ 45.7k

an open source, extensible AI agent that goes beyond code suggestions - install, execute, edit, and test with any LLM

Antigravity Awesome Skills

★ 38.3k

Installable GitHub library of 1,400+ agentic skills for Claude Code, Cursor, Codex CLI, Gemini CLI, Antigravity, and more. Includes installer CLI, bundles, workflows, and official/community skill collections.

AgentScope

★ 25.5k

Build and run agents you can see, understand and trust.

Serena

★ 24.5k

A coding agent toolkit that provides IDE-like semantic code retrieval and editing tools, enabling LLMs to efficiently navigate and modify codebases using symbol-level operations instead of basic file reading and string replacements.

Browse More Coding Agents MCP Servers

Explore all coding agents servers available in the MCPgee directory. Each server includes setup guides for Claude, Cursor, and VS Code.

Coding AgentsBrowse All Servers

Quick Config Preview

{ "mcpServers": { "mai-ui": { "command": "npx", "args": ["-y", "mai-ui"] } } }

Add this to your claude_desktop_config.json or .cursor/mcp.json

Read the full setup guide →

Ready to use MAI UI?

Browse our complete directory of 33,000+ MCP servers, read setup guides for your editor, and start building with the Model Context Protocol.

33,000+ ServersFree & Open SourceStep-by-Step Guides
Explore All ServersRead Our Guides