Open Minis

Product Introduction

1. Definition: Open Minis is a free, open-source, on-device AI agent application for iOS and Android. Technically, it is a mobile-native framework that provides a large language model (LLM) with a secure, sandboxed execution environment and native system-level integrations.
2. Core Value Proposition: It exists to provide a truly private, powerful, and versatile AI assistant that operates entirely on your smartphone. Its primary value is enabling AI models to perform real computational tasks—like coding, file management, and web browsing—and interact with native device data (Health, Calendar, HomeKit) without ever sending sensitive keys or personal information to the cloud.

Main Features

1. On-Device Alpine Linux Sandbox: At its core, Open Minis runs a lightweight, sandboxed Alpine Linux virtual environment directly on your mobile device. This provides the AI agent with a real, isolated computer to work within. It can install packages via apk, execute Python scripts, edit local files, and run command-line tools, all without an internet connection or external servers.
2. Bring-Your-Own-Model (BYOM) Flexibility: The app is model-agnostic. Users can connect it to virtually any major AI API endpoint, including Anthropic Claude, OpenAI GPT, Google Gemini, xAI Grok, Kimi, or any OpenAI-compatible service via OpenRouter. This allows users to leverage the specific strengths of different LLMs while maintaining data privacy.
3. Native iOS/Android Bridge System: Open Minis includes over 30 native bridges that allow the AI agent to securely interact with the host operating system's frameworks. This includes deep integrations with HealthKit (read health data), HomeKit (control smart home devices), Calendar & Reminders (CRUD operations), Photos (access media library), Contacts, Bluetooth, Clipboard, and Device sensors, enabling context-aware automation.
4. Integrated Privacy-First Browser: The agent has direct access to a built-in, sandboxed web browser (minis://browser/). This allows the AI to perform web research, extract information from websites, automate form interactions, or control web applications, all within the app's secure environment without exposing browsing data.
5. Local File System & Workspace Management: The app provides structured local filesystem access via dedicated URI schemes like minis://workspace/ for project files and minis://attachments/ for user-uploaded documents. This turns the phone into a portable development and automation workstation where the AI can persistently manage and edit real files.

Problems Solved

1. Pain Point: The privacy risk and data sovereignty concerns of cloud-based AI assistants. Traditional AI chatbots process all requests on remote servers, meaning sensitive queries, personal data, and API keys are transmitted and stored externally.
2. Pain Point: The limited functionality of most AI assistants. Standard chatbots are confined to text generation and lack the ability to execute code, manage local files, or interact with other apps and system services, severely limiting their utility for automation and complex tasks.
3. Target Audience: Privacy-conscious power users, developers, and tinkerers who want to automate mobile workflows. Technical personas include iOS/Android developers testing scripts, data analysts running Python on the go, and smart home enthusiasts creating complex HomeKit automations.
4. Use Cases: A developer can ask the agent to write a Python script to analyze a local CSV file (minis://attachments/data.csv) and generate a chart. A user can ask it to scan their HealthKit data to summarize weekly exercise trends. Another can instruct it to browse a product website, extract specifications, and save them to a note in their Reminders app—all executed locally.

Unique Advantages

1. Differentiation: Unlike cloud-only AI assistants (ChatGPT, Copilot) or device-limited ones (Siri, Google Assistant), Open Minis combines the power of a cloud LLM's reasoning with the privacy and system access of a local agent. Compared to other automation apps, it uses natural language as the interface instead of requiring manual block-based programming.
2. Key Innovation: The architecture of pairing a sandboxed Linux shell with a native bridge system. This dual-layer approach is unique. The Linux shell handles generic computation and scripting, while the native bridges provide safe, sanctioned access to sensitive mobile OS features, creating a balance between capability and security not found in other apps.

Frequently Asked Questions (FAQ)

1. Is Open Minis really free and private? Yes, Open Minis is completely free, open-source (FOSS), and architected for privacy. Your API keys, conversation history, and all data processed through the Linux sandbox or native bridges remain on your physical device and are never transmitted to the developers' servers.
2. What can the Open Minis AI agent actually do on my phone? The AI agent can perform computational tasks in its Linux environment (coding, math, text processing), manage files in its workspace, control a private browser, and read/write data to your Health, Calendar, Reminders, Contacts, and HomeKit apps based on your direct commands and permissions.
3. Do I need coding skills to use Open Minis? Basic technical comfort is helpful, but deep coding skills are not required. The primary interface is natural language conversation with your chosen AI model. You instruct it in plain English (or other languages) to perform tasks, though understanding concepts like files, scripts, and APIs will unlock its full potential.
4. How does the on-device Alpine Linux shell work without internet? The Alpine Linux distribution is extremely lightweight and runs in a local container (sandbox) on your phone. Once the app is installed, this environment is fully local. It can execute installed packages and scripts offline; internet is only required if the AI model itself needs to call its cloud API or if you ask it to install a new package or use the browser.