Vesta AI Explorer

Vesta AI Explorer is a native macOS application designed to provide an agentic AI chat experience with a focus on privacy and on-device processing for Apple Silicon users. It offers a natural language interface to Vesta, enabling users to interact with multiple AI backends, including Apple Intelligence, MLX, llama.cpp, OpenAI, and HuggingFace. The application supports seamless switching between these backends, each optimized for different use cases, and allows all inference to run locally on Apple Silicon's Neural Engine when public AI access is disabled.

The platform includes a range of features such as vision analysis using local Vision Language Models (VLMs), LaTeX rendering via KaTeX for mathematical equations, and code blocks with syntax highlighting for over 20 programming languages through highlight.js integration. Users can analyze images, diagrams, and documents with local VLMs, utilize drag-and-drop or paste functionality for image input, and view inline image displays with rich markdown descriptions. The application also supports creative writing with rich formatting, image generation with FLUX and Stable Diffusion, and video creation with Wan2.2 and HunyuanVideo via HuggingFace.

For automation and workflow integration, Vesta AI Explorer exposes over 33 tools via the Model Context Protocol (MCP), allowing remote control and scripting from MCP-compatible clients such as Claude Code. The in-app model browser enables users to explore and download models directly from HuggingFace, and the settings panel provides controls for model management, generation parameters, and TTS configuration. Text-to-speech is handled by Kokoro with over 45 voices, and speech-to-text is supported through WhisperKit, both operating on-device via MLX and CoreML.

Vesta AI Explorer can connect to external AI providers through APIs, supporting 31 pre-configured providers including OpenAI, OpenRouter, Cerebras, Groq, and others, with access to over 115 models across various endpoints. The application is built with SwiftUI for macOS 15+ and requires macOS 26 or later on Apple Silicon (M1, M2, M3, M4). Installation options include downloading a DMG or using Homebrew. It is tailored for users seeking a private, flexible, and multi-modal AI assistant on their Mac.