heshengtao/super-agent-party

What it solves

Super Agent Party is an AI desktop companion that transforms AI agents into interactive, visual avatars. It solves the problem of static chat interfaces by providing a highly customizable 3D/2D visual presence (VRM/Live2D) and enabling agents to interact directly with the user's computer, messaging platforms, and live streaming services.

How it works

The project integrates multiple AI models (reasoning, vision, speech, and image generation) with a desktop application. It uses a toolchain for mouse, keyboard, and terminal control to allow agents to perform tasks on the computer. It also supports the Model Context Protocol (MCP) for external tool connectivity and provides an OpenAI-compatible API for developer integration.

Who it’s for

• End users who want a visual AI companion with customizable avatars and immersive chat experiences.
• Content creators/VTubers who want to integrate AI agents into live streams (Bilibili, YouTube, Twitch) or control Vtube Studio models.
• Developers who want to build AI agents with computer control capabilities and an extensible plugin system.

Highlights

• Visual Avatars: Supports custom VRM models, THA-based 2D characters, and Live2D integration via Vtube Studio.
• Computer Control: AI agents can use desktop vision to control the mouse, keyboard, and terminal to complete tasks.
• Multi-Platform Deployment: One-click deployment to IM bots (QQ, WeChat, Telegram, Discord, Slack) and live streaming platforms.
• Extensible System: A robust plugin architecture allowing users to install or create new extensions (e.g., AI Editor, AI PPT, AI Tarot reader).
• Developer Friendly: Open OpenAI-compatible API and MCP interfaces for external connectivity.