Reins is a private, local AI chat app for iPhone, iPad and Mac that connects to Ollama, LM Studio or any OpenAI-compatible server, or runs MLX and Apple Foundation Models on-device, and it is built for Apple users who want LLM conversations without an account, without cloud uploads and without data collection.

What it is

Reins is a chat client written in Dart that runs on iOS, iPadOS and macOS and sits at the front of the local LLM stack. Rather than hosting a model itself, it acts as a single interface over the inference options an Apple user already has: a self-hosted Ollama or LM Studio instance, a remote OpenAI-compatible endpoint, Ollama Cloud, or models executed directly on the device through MLX and Apple's Foundation Models. Models loaded on-device can also be shared with other devices on the local network, so one machine can serve a household or an office. Each conversation carries its own model, system prompt, temperature and context size, and the app manages model downloads and multiple server connections from within its own interface.

The concrete problem it solves is the login-and-telemetry model of hosted AI chat. Reins has no login and performs no data collection, and chats remain on the device, so it replaces the account-based hosted chat service for people who already run their own inference. It also removes the friction of juggling separate tools for model management, server switching and prompt configuration: adding a server with its API key and custom headers, downloading a model, and setting a per-chat system prompt all happen in the same app. Tool calling, web search and page fetching work without an API key, so the capability that usually requires a paid cloud account is available against a locally served model.

Key capabilities

• Connects to Ollama, LM Studio, Ollama Cloud or any OpenAI-compatible server, with multiple servers supported, each with its own API keys and custom headers and instant switching between them.
• Runs MLX models and Apple's Foundation Models on-device with no server at all, and shares those on-device models with other devices on the local network.
• Gives models tools: tool calling, web search, page fetching and multi-step reasoning, with no API key required.
• Applies per-chat configuration for system prompt, model, temperature and context size, so every conversation can be tuned independently.
• Manages models and servers in-app, including browsing, downloading and managing models.
• Accepts attachments as images, PDFs, CSVs and text files.
• Supports branching and resumption: duplicate a chat to explore another path, and continue a cut-off response from where it stopped.
• Adds thinking mode for harder questions, temporary chats that leave no trace, export and import for backup or sharing, and background generation that keeps running after leaving the app.

Who uses it and how

• Apple-only users who run Ollama or LM Studio on a home machine or LAN server and want a native iPhone, iPad and macOS front end instead of a browser tab.
• Privacy-focused users who need chats to stay on the device and refuse logins or telemetry, using temporary chats when nothing should persist.
• People running on-device MLX or Foundation Models with no server, who then share the loaded model with other devices on the same local network.
• Users whose models need web access but who have no API key, relying on the built-in tools, web search and page fetching for multi-step answers.
• Users who work across several backends, keeping multiple servers configured with separate keys and headers and switching instantly per chat.

Getting started

The README excerpt gives no install or build command: it states only that the repository holds the source code of Reins up to version 1.3.4, with the website at https://getreins.app as the product entry point and GitHub issues plus support@getreins.app for problems. Anyone building from source needs a Dart and Flutter toolchain for Apple targets, since no Docker image, package name or compose file is provided.

How it compares

No list of paid alternatives is given in these facts, so this is positioning rather than a head-to-head. Reins is a client in the local LLM ecosystem, not a competitor to Ollama, LM Studio, Ollama Cloud or OpenAI-compatible servers: it speaks to them, and it interoperates with MLX and Apple's Foundation Models rather than replacing them. Other chat front ends for these backends exist in the wider world, but none is named in this registry entry.

When to use it — and when not to

Pick Reins if the devices are Apple, the models are already served locally or run on-device, and the priority is that no account exists and no chat leaves the machine. Be aware that the README excerpt documents no build or install path, so self-hosters must supply their own Flutter toolchain and assemble the app, and the repository is explicitly frozen at version 1.3.4 while the shipped product has moved on. Skip it if you need Android, Windows or Linux clients, or if you want a packaged binary you can install without developer tooling; the licence is GPL-3.0, which matters if you intend to redistribute a modified build.