Caverno vs Enigmus

Caverno is a powerful chat client for locally-hosted LLMs. Connect to Ollama, LM Studio, vLLM, or any OpenAI-compatible API endpoint and start chatting — no cloud required. KEY FEATURES • Any OpenAI-Compatible Server Connect to your own LLM running on localhost or your local network. Supports any endpoint that speaks the OpenAI API format. • Tool Calling (MCP Protocol) Extend your AI with real-time tools — web search, calculator, code execution, and more. Caverno uses the Model Context Protocol to let your LLM take actions. • Session Memory Your assistant remembers your preferences, constraints, and context across conversations. Memory is stored locally and fully under your control. • Voice In & Out Speak your messages and have responses read aloud. Configurable speech rate and auto-read options. • Image Attachments Attach photos from your library to include visual context in your conversations. • Multiple Conversations Create, switch, and manage separate chat threads. All conversations are persisted locally. • Assistant Modes Switch between General and Coding modes for different response styles tailored to your task. • Privacy First No account required. No data sent to third parties. All conversations and memory stay on your device — the only network connection is to the server you configure. GETTING STARTED 1. Install an OpenAI-compatible LLM server (Ollama, LM Studio, etc.) 2. Open Caverno and go to Settings 3. Enter your server URL and select a model 4. Start chatting! Caverno is designed for developers, AI enthusiasts, and anyone who wants full control over their AI assistant.

Enigmus runs large language models entirely on your device. Conversations never leave your iPhone, iPad, or Mac - no accounts, no servers, no network required after the initial model download. Hardware requirements: iPhone 13 or newer (A15 Bionic or later), an iPad with an M1 chip or newer, or any Mac with Apple silicon. Features - Ships with a small bundled model, so you can start chatting on first launch without downloading anything. - Download additional open-weight models from Hugging Face, including Qwen3.5/3.6 and Google's Gemma 4 families. The picker recommends models that fit your device's available memory. - Streaming responses with markdown and code-block rendering. - Reasoning-capable models show their thinking before the final answer. - Local chat history backed by SwiftData — conversations stay on the device. - Adaptive layout for iPhone, iPad, and Mac, with keyboard and pointer support on the larger screens. - Customize tint color, fonts, and haptics in Settings. Privacy All inference runs through MLX, Apple's on-device machine-learning framework. No telemetry, no analytics, no account. The only outbound traffic is the optional model download from Hugging Face's public CDN.