RecurseChat

RecurseChat is a Mac-native application designed for users who want to interact with artificial intelligence models directly on their own device, without relying on cloud servers or internet connectivity. The core functionality centers on running large language models (LLMs) locally on a Mac computer, enabling private, offline, and fast conversations with AI. Users can chat with local models such as GGUF-format models from Hugging Face, as well as connect to cloud-based providers like Anthropic’s Claude and OpenAI’s ChatGPT from within a single application interface. No subscription is required; the app operates on a one-time purchase model with a free trial available. The application emphasizes local-first architecture, meaning all AI computations happen on the user’s machine. This design ensures that no conversation data ever leaves the device, addressing privacy concerns for sensitive or confidential discussions. Users can run models entirely offline, bypassing the need for internet access or external servers. The app supports a wide range of local models, allowing users to choose from various sizes and architectures depending on their hardware capabilities and preference for speed versus reasoning depth. A standout feature is the floating chat window, which can be summoned from any application using a keyboard shortcut (⌘+⇧+Enter). This overlay appears on top of all other windows, enabling users to ask questions, get explanations, or generate text without switching contexts. The floating chat remains persistent and can be resized or hidden as needed, making it suitable for workflows that require rapid access to AI assistance while coding, writing, or researching. RecurseChat also supports file-based interactions through local Retrieval-Augmented Generation (RAG). Us