An Android chat client that doesn't care where your model lives

What it does Maid is a React Native Android app that acts as a unified front-end for LLMs. It runs GGUF models locally through llama.cpp when you want offline access, and connects to remote providers like OpenAI, Anthropic, DeepSeek, and Ollama when you have an API key. You can download curated models straight from Hugging Face, load your own GGUF files from storage, and manage conversations as portable JSON.

The interesting bit Most mobile AI clients pick a lane—either local tinkerer or cloud consumer. Maid sits in both camps, letting you switch between a downloaded Hugging Face model and a remote endpoint inside the same conversation list. The app also bundles a curated model browser, which turns the hunt for compatible GGUF files into something closer to a one-tap install.

Key highlights

• Fully on-device inference via llama.cpp with no network required
• Remote support for Anthropic, DeepSeek, Mistral, Novita, Ollama, and OpenAI
• One-tap downloads of curated Hugging Face models including Qwen, Phi, and TinyLlama
• Chat export and import as JSON for moving conversations between devices
• Optional Supabase account sync to back up settings and history
• MIT licensed, no telemetry, no ads, and Material You theming

Verdict Worth installing if you want a unified, open-source Android inbox for both offline local models and paid cloud APIs. Skip it if you need an iOS or desktop client.