AetherLink: An Open-Source Cross-Platform AI Assistant Solution Supporting Multi-Models and MCP Protocol

AetherLink is an open-source cross-platform AI assistant application built with React, Capacitor, and Tauri, designed to address the fragmentation issue of AI clients. It supports mainstream models like OpenAI, Claude, and Gemini, integrates MCP protocol extension capabilities, and provides features such as voice interaction, knowledge base management, and code highlighting, offering a unified and scalable AI dialogue client solution for individuals and small teams.

With the development of large language models, the AI dialogue client market is thriving, but most solutions face challenges such as single functionality, platform limitations, or insufficient scalability. Commercial products are locked into specific ecosystems, while open-source projects are often restricted by their tech stacks and struggle to cover both mobile and desktop platforms. Users find it hard to find a client that allows flexible model switching and provides a consistent experience across devices.

• Multi-model Support: Covers OpenAI (GPT series), Anthropic Claude (3/3.5 series), Google Gemini (1.5 Pro/Flash), xAI Grok, SiliconFlow, Volcengine, and custom API endpoints. Users can flexibly configure keys and switch models.
• MCP Protocol Integration: Supports connecting to external data sources, calling third-party tools, and enabling deep integration between AI and the local environment, going beyond a simple dialogue client.
• Voice Interaction: Speech recognition supports OpenAI Whisper, Capacitor Native, and Web Speech API; speech synthesis supports Azure TTS, OpenAI TTS, and SiliconFlow TTS.
• Knowledge Base Management: Supports uploading documents like PDF/Word/TXT, semantic search based on vector embeddings, and AI answers can reference knowledge base content.

• Frontend Stack: React 19.x, TypeScript 5.9+, Material UI 7.x, Vite 8.x, SWC, Redux Toolkit + Zustand, Tailwind CSS + Emotion.
• Cross-Platform Solution: Capacitor (for Android/iOS mobile) and Tauri (for desktop, small size and low memory usage).
• Data Storage: IndexedDB (chat history/knowledge base), LocalStorage (user settings), and file system (document import/export).
• AI SDK: Vercel AI SDK (unified streaming processing) + OpenAI SDK (direct API calls).

• Personal AI Assistant: Users can choose models based on tasks (Claude for complex reasoning, GPT-4o for creative writing, Gemini 1.5 Pro for long documents).
• Developer Tool: Connect to local code repositories via MCP to enable AI to understand project context and provide precise suggestions.
• Enterprise Knowledge Management: Integrate enterprise documents, allowing employees to retrieve information via natural language queries.

Current Limitations: 1. Insufficient support for local models (Ollama, llama.cpp); 2. Lack of multi-user collaboration and dialogue sharing; 3. No MCP plugin market established; 4. Performance on low-end devices needs optimization. Improvement Directions: Enhance local model support, develop collaboration features, build a plugin ecosystem, and optimize performance on low-end devices.

AetherLink provides a unified access layer, cross-platform experience, and scalable capability platform, helping users break free from single-service provider lock-in. Open-Source License: AGPL v3.0 for individuals/small teams (≤8 people), commercial license required for enterprises. Community Support: GitHub Issues, QQ Group (930126592), contribution guidelines. The project has been released on GitHub, with complete documentation and build guides.