Vesta AI Explorer — AI Chat for macOS

Everything you need, on your Mac



Apple Intelligence

Built-in Foundation Models framework. Completely private, zero setup, runs on Neural Engine.

⚡

MLX & llama.cpp

Run MLX and GGUF models locally. Explore and download them directly from HuggingFace with an in-app model browser.

👁

Vision & VLM

Analyze images, diagrams, documents with local Vision Language Models. Drag-and-drop or paste.

🎤

TTS & STT

Kokoro text-to-speech with 45+ voices. WhisperKit speech-to-text. All on-device via MLX & CoreML.

🖼

Image & Video Gen

Generate images with FLUX and Stable Diffusion. Create videos with Wan2.2 and HunyuanVideo via HuggingFace.

🔗

MCP Protocol

33+ tools via Model Context Protocol. Automate Vesta from Claude Code, scripts, or any MCP client.

Beyond On-Device

Access Public AI Models

Connect to the broader AI ecosystem through HuggingFace Inference API and OpenAI-compatible providers like OpenRouter, Cerebras, Groq, Together AI, and more — with 31 providers pre-configured out of the box.

HuggingFace

OpenAI

OpenRouter

Cerebras

Groq + 26 more providers

Model Context Protocol

✨ Agentic Sidekick

Vesta exposes 33+ tools over TCP using the Model Context Protocol. Automate workflows from Claude Code, scripts, or any MCP-compatible client.

Watch MCP in action

See it in action

Live captures driven via MCP remote control

Click to enlarge

UIWelcome

Welcome to Vesta AI Explorer

Clean, dark-themed native macOS interface with conversation sidebar, backend selector, and integrated settings panel. Built with SwiftUI for macOS 15+.

Click to enlarge

Apple IntelligenceChat

Apple Intelligence On-Device

Chat powered by Apple's Foundation Models framework. Fully private, running entirely on Apple Silicon's Neural Engine.

55 tok/s on-device

Click to enlarge

CodeSyntax Highlighting

Syntax-Highlighted Code Blocks

Full highlight.js integration with 20+ languages. Swift, Python, JavaScript, Rust, and more — rendered beautifully with copy support.

Click to enlarge

MLXCodeScience

Local LLM via MLX (84 tok/s)

Qwen3-VL-4B running locally on Apple Silicon via MLX framework. Beautiful analogy + Python simulation with syntax highlighting. No internet required.

84 tok/s Apple Silicon

Click to enlarge

VisionVLM

Vision Language Model (VLM)

Analyze images with local VLMs. Inline image display with rich markdown descriptions — breed identification, scene analysis, and more.

$LaTeX Math Rendering$

Click to enlarge

LaTeXKaTeX

Beautiful Math with KaTeX

Full LaTeX math rendering via KaTeX. Display and inline equations, Taylor series, matrices, integrals — publication-quality typography.

Click to enlarge

TablesGFM

GitHub Flavored Markdown Tables

Full GFM support with styled tables, strikethrough, task lists, and more. Combined with LaTeX math in the same conversation.

Click to enlarge

CreativeMLX

Creative Writing + Code

Cyberpunk microfiction with rich formatting — headers, blockquotes, inline code, and a Swift greeting function. All from a local 4-bit model.

Click to enlarge

VisionArchitectureAWS

Technical Diagram Analysis

VLM analyzes a complex AWS 3-tier architecture diagram — identifying services, tiers, and infrastructure components from a single screenshot.

Click to enlarge

VisionRecipes

Recipe Recognition

Point the VLM at a cookbook page — it identifies the recipe, lists ingredients, and suggests healthier modifications. Practical AI for everyday use.

Click to enlarge

Apple IntelligenceBackend

Seamless Backend Switching

Switch between Apple Intelligence, MLX, llama.cpp, OpenAI, and HuggingFace with a single click. Each backend optimized for different use cases.

Click to enlarge

MLXBackend

MLX: Apple Silicon Optimized

Run Qwen, Llama, Gemma, Phi, and more locally. Unified memory, Metal GPU acceleration, 4-bit quantization. Your data never leaves your Mac.

84 tok/s Qwen3-VL-4B

Click to enlarge

MLXSettings

MLX Settings

Model management, temperature, top-p, repetition penalty, token limits, and TTS configuration — all in one panel.

Click to enlarge

llama.cppSettings

llama.cpp Settings

GGUF model library, vision/text mode toggle, generation parameters, context size, and WhisperKit STT configuration.

Click to enlarge

External AISettings

External AI by API

Connect to OpenAI, OpenRouter, Cerebras, Groq, and 26+ providers. Search 115+ models across US, Canadian, and international endpoints.

Click to enlarge

HuggingFaceSettings

HuggingFace Explorer

Chat, Vision, Image generation, Editing, Transcription, and Video — six tabs of cloud AI capabilities with auto provider selection.

Click to enlarge

All BackendsSettings

Backend & Model Explorer

The unified settings inspector — toggle backends, load models, adjust parameters, and see status at a glance across all five engines.

Click to enlarge

HuggingFaceBackend

HuggingFace: Cloud AI Hub

Quick-access pills for Image generation, Video, Vision, Search, Software Maps, and more — one-click access to cloud AI capabilities.