RecurseChat
RecurseChat is a personal Artificial Intelligence application that enables users to engage in conversations with local AI models. Primarily...
Last verified:
What is RecurseChat?
RecurseChat is a personal AI application exclusively for Mac users with Apple Silicon, enabling private and offline chats with local large language models (LLMs) alongside integration with cloud services like Claude and ChatGPT. It emphasizes data ownership, allowing users to import ChatGPT history, continue conversations locally, and export chats anytime without vendor lock-in. The app runs entirely within the Mac App Sandbox for enhanced security, ensuring conversations with local models never leave the device and no telemetry is collected.
Key features include chatting with local documents such as PDFs, markdown, and text files using on-device RAG with embedding and vector search; multimodal image interactions; fast full-text search across thousands of messages; quick navigation through history; multi-model usage in single sessions; chat organization via bookmarks and nested folders; message editing, branching, and switching; and customizable AI personalities and interface settings. A floating chat window summonable via ⌘+⇧+Enter works across any app, providing instant access.
Designed for users seeking a seamless daily AI companion, RecurseChat supports zero-config setup for models like Meta Llama 3.1/3.2/3, Gemma 2/3, Mistral, LLaVA, DeepSeek R1, and custom GGUF files from Hugging Face or Ollama. It also handles OpenAI-compatible endpoints. Ideal for privacy-conscious individuals, AI enthusiasts, and professionals needing offline document analysis or a unified client for multiple AI providers without subscriptions.
RecurseChat pricing
Pricing model: Paid
One-time purchase via Mac App Store. No subscription required. No free tier mentioned. No data collection.
RecurseChat pros
- Runs local LLMs offline on Mac
- Zero internet needed for local chats
- Chats stay private on device
- No app subscription required
- One-time purchase model
- Imports ChatGPT history seamlessly
- Fast full-text search on thousands of messages
- Chats with PDFs offline via local RAG
- Multimodal image interaction support
- Floating chat via ⌘+⇧+Enter shortcut
- Multi-model use in same session
- Bookmarks and nested folders for organization
- Message editing and branching
- Supports Llama 3.1/3.2/3, Gemma 2/3
- Custom GGUF and Hugging Face models
- Ollama model import capability
- Claude 3.5 and GPT-4o integration
- Mac App Sandbox for security
- No usage telemetry collected
- Zero-config model download
RecurseChat cons
- Mac with Apple Silicon only
- macOS Ventura 13.5 or later required
- Recommends 16GB RAM minimum
- CPU fallback on 8GB RAM slows inference
- No Windows or Linux support planned soon
- Online models need own API keys
- GPT-4 requires OpenAI $5 payment tier
- Limited to Apple ecosystem
- No iOS or iPad version mentioned
Frequently asked questions about RecurseChat
What’s the system requirement? Will RecurseChat be available on other Operating Systems?
Supports Mac with Apple Silicon on macOS Ventura 13.5 or later. Recommends 16GB RAM or more; on 8GB uses CPU inference for default models. No immediate plans for other OS, but subscribe for updates.
How do I start using RecurseChat?
Step 1: Download from Mac App Store. Step 2: Type in chat box; app prompts model download and auto-starts local AI server. Step 3: No further steps needed.
What AI models does RecurseChat support?
Local: Meta Llama 3.2/3.1/3, Gemma 3/2, DeepSeek R1, Mistral, Llava, WizardCoder Python, custom GGUF via llama.cpp, Ollama imports. Online: GPT-4.1/4o/4o mini, Claude 3.5/3 (needs API keys). OpenAI-compatible endpoints like OpenRouter.
How does RecurseChat compare to other similar applications?
Feels like online AI but with local ownership: import/export ChatGPT history, intuitive UX for beginners with advanced customizations, fully private local chats never leaving machine, no vendor lock-in.
How private and secure is RecurseChat?
No usage or telemetry collected. Runs in Mac App Sandbox limiting access to selected files/network. Local chats stay on-device; online subject to provider policies like OpenAI.
Can I chat with documents in RecurseChat?
Yes, chat with PDFs, markdown, txt files completely offline using local RAG with on-device embedding and vector search. Files never leave your Mac.
Does RecurseChat support image chats?
Yes, interact with images using multimodal chat, powered by models like LLaVA for vision capabilities.
How do I import ChatGPT history?
Export from ChatGPT settings as conversations.json, then in RecurseChat Settings > Chat History > Import, select file. Becomes searchable and continuable with local AI.
What is the floating chat feature?
Press ⌘+⇧+Enter to summon AI chat window always on top of any app, enabling instant access without switching applications.
Can I use multiple models in one chat?
Yes, switch and use multiple AI models within the same chat session for flexible interactions.