Open Minis Goes Fully Open Source, Democratizing Smartphone AI Agents

The developer, Ethan, announced on July 25, 2026, that the full codebase of the smartphone AI agent “Open Minis” will be made open source. Compatible with both iOS and Android platforms, this application has undergone three to four months of iterative development and is now used by tens of thousands of users. According to reports by the niche tech blog Small Blue Frog, Open Minis has been described as “probably the best agent application for your smartphone,” boasting features such as a fully local Alpine Linux environment, browser automation, expandable skills, persistent memory, and deep system integration.

The Background of Going Open Source

In the announcement, Ethan stated that the complete code for both the iOS and Android versions of Open Minis is now publicly available. The application has a proven track record of supporting the daily workflows of thousands of users. Since its early development, the project has envisioned itself as a “true AI agent on your device,” and it is now transitioning to community-driven development.

Some have dubbed this product the “smartphone version of OpenClaw.” While OpenClaw operates PCs, Open Minis directly controls smartphones. It can also be seen as a third-party equivalent of Apple Intelligence—a Siri-like assistant that actually performs tasks. Unlike traditional virtual assistants that limit themselves to providing information or handling simple tasks, Open Minis is designed as an agent capable of operating the device itself.

An Agent with Memory

One of the most noteworthy features of Open Minis is its memory functionality. The application can read past chat history and refer to it when executing new tasks. In other words, Open Minis has a persistent memory that allows it to consider user behavior patterns and past instructions when carrying out actions.

This memory functionality is not merely a history log but serves as contextual information for the agent to make autonomous decisions. For instance, if the user previously expressed a preference for a specific data format in a conversation, Open Minis would generate output reflecting that preference. This continuous learning and adaptation mechanism is a unique feature not found in traditional virtual assistants.

The Power of the Embedded Alpine Linux

Environment

Open Minis includes a fully local Alpine Linux environment. Within this environment, its AI agent can autonomously install programs, write code, and process data. For example, video processing via ffmpeg or content downloading using yt-dlp is possible.

This design aligns with the direction outlined in a recent article about OpenAI’s release of the Sol, Terra, and Luna models in GPT-5.6, which emphasized “AI agents with code execution capabilities.” By not relying on cloud-based virtual environments and completing all processing locally on the device, Open Minis offers advantages in both privacy and response speed.

The native Linux shell environment provides virtually limitless extensibility for the AI agent. Users can add custom scripts and leverage existing Linux tools directly. The full potential of this design becomes evident when considering workflows like “automatically analyzing Jenkins error logs with n8n and Claude and sending Slack notifications,” which shows the capability for comprehensive automation directly on a smartphone.

Browser Automation and Web Interaction

Open Minis comes with a built-in browser that allows the AI agent to autonomously browse web pages, fill out forms, extract content, and capture screenshots. All these operations are conducted entirely within the Open Minis application and are fully automated.

This feature enables users to delegate tasks such as information gathering and data entry without manually operating apps. For example, the AI could collect data from specific websites periodically and save it in a specified format, all triggered by a single verbal command. The built-in browser provides functionalities equivalent to regular mobile browsers, including cookie management and session retention.

Depth of System Integration

Open Minis has extensive access to iOS and Android system permissions. Major data sources on the device, such as calendars, contacts, location information, photos, clipboard, weather, and notifications, are accessible. The iOS version additionally supports HealthKit, HomeKit, reminders, Bluetooth, and NFC, while the Android version offers automation through accessibility services and supports Shizuku privilege operations.

This deep system integration enables Open Minis to go beyond simple information retrieval and actively operate the device. For instance, compared to Apple’s New Siri, expected to debut in iOS 27 as a “true assistant,” Open Minis stands out as a third-party application achieving comparable or greater system access. Open Minis implements features that Apple restricts within its own platform, showcasing the power of an open-source AI agent.

Diverse Practical Applications

Open Minis can execute the following practical tasks through simple language commands:

• Automatically create schedules by sharing calendar content without opening the calendar app.
• Analyze health data directly to assess the user’s condition.
• Evaluate heart health using Apple Watch data.
• Search for, skip, and play songs on Spotify through voice commands without opening the app.
• Edit videos from raw footage to final production in one step.
• Organize music found on TikTok into YouTube Music playlists with a single tap.
• Set morning alarms that simultaneously read out the news.
• Detect important content in group messages and add it to reminders.
• Automatically organize notes from web pages opened.
• Record caffeine intake automatically by photographing coffee.
• Set complex alarms and reminders in bulk.

These use cases demonstrate that AI agents can function as practical tools in everyday life, rather than just as chatbots. Tasks that previously required multiple apps and manual steps can now be completed in a single interaction.

How to Access and Use

The source code for Open Minis is available on its official website and GitHub. However, users need to provide their own AI APIs, such as OpenAI’s API. The application itself is free to use, and the code is fully open source. It does not include any built-in models; all processing is performed through the user’s chosen API.

This approach offers the flexibility for users to select AI models that best fit their needs, whether it’s OpenAI’s GPT series, Anthropic’s Claude, or locally operated open-weight models. Users can adjust response quality, cost, and privacy levels according to their preferences, making it appealing to those with high technical literacy.

Differences from Existing AI Assistants

Unlike Apple’s Siri or Google Assistant, which primarily provide information and perform basic device controls, Open Minis is designed as a full-fledged task-executing agent. Its built-in Linux environment and browser automation capabilities set it apart from existing virtual assistants.

Additionally, the open-source nature of Open Minis is a significant advantage. With its publicly available code, security audits are possible, and the community can contribute to feature expansions. Users can fork the code to build their own custom agents tailored to their specific needs.

Editorial Opinion

In the short term, the open-sourcing of Open Minis is expected to accelerate the development of smartphone AI agents. Particularly in the Android ecosystem, where support for Shizuku privilege operations has already been realized, advanced automation without requiring root access is now possible. On iOS, integration with HealthKit and HomeKit is likely to create competition with Apple’s own implementations. Over the next three to six months, we can expect multiple derivative projects and commercial services based on this codebase to emerge.

In the long term, AI agents on devices have the potential to revolutionize personal computing. Just as OpenClaw demonstrated the automation of PC operations, Open Minis showcases similar capabilities on smartphones, which are ubiquitous and always carried by users. Over the next one to three years, we may see a significant shift from cloud-dependent AI assistants to decentralized, on-device AI agents. With the freedom to choose APIs and the enhanced focus on privacy, this trend is likely to appeal to enterprise users as well.

References

• “Open Minis 宣布开源：比 Siri 更像 AI 助手，让手机帮你完成任务。支持 iOS、Android”, by 青小蛙 — 小众软件, 2026-07-26T09:31:06.000Z (ARR)
• Source URL: https://www.appinn.com/open-minis-open-source/