Embed an on device AI agent inside a Flutter mobile app
Let an agent build, update, and install an Android app from the phone itself
Expose sandboxed file tools like image compression to an agent
Connect an agent to smart home or other external devices
| antgroup/napaxi | alexgladkov/quickai | criccomini/sleet | |
|---|---|---|---|
| Stars | 26 | 26 | 26 |
| Language | Rust | Rust | Rust |
| Setup difficulty | hard | easy | moderate |
| Complexity | 4/5 | 2/5 | 4/5 |
| Audience | developer | developer | developer |
Figures from each repo's GitHub metadata at analysis time.
Requires Rust, Git LFS, and native Android or iOS build tools depending on target platform.
Napaxi is a software development kit that lets app developers add an AI agent directly inside their mobile app, running on the phone itself rather than through a cloud server. The core engine is written in Rust and shared across different mobile platforms, with a Flutter version being the most complete right now, alongside adapters for Android and iOS being developed alongside it. The main idea is that the agent's memory, session history, files, and tool use all stay on the device, so nothing needs to be sent to a Napaxi run server beyond whatever calls the host app itself chooses to make to an AI model. Any tool the agent wants to use, such as reading a file, opening a browser, or controlling a smart home device, runs inside a sandboxed environment where the surrounding app decides exactly what permissions to grant. Developers building on Napaxi keep full control of their app's look, accounts, and which AI model or provider it connects to. Napaxi supplies the reusable pieces underneath: session handling, stored workspace state, a set of built in tools, support for the Model Context Protocol, and connections to other apps, agents, and devices, described in the project as xApp, xAgent, and xChannel connectivity. Example uses shown in the project include building and installing an Android app entirely from the phone, compressing an image through a sandboxed file tool, and controlling a smart home light through a connected device provider. Getting started requires the Rust toolchain, the Flutter SDK if working with that adapter, Git LFS for downloading bundled binary assets, and Android or iOS native build tools depending on target platform. The project is released under the GPL 3.0 or later license, meaning any project built using this code and distributed to others must also share its own source code under compatible terms.
A mobile SDK that lets developers embed an on device AI agent directly inside their app, keeping data and tool execution on the phone instead of a cloud server.
Mainly Rust. The stack also includes Rust, Flutter, Kotlin.
GPL 3.0 or later requires that any app or project using this code, if distributed, must also share its source code under the same license.
Setup difficulty is rated hard, with roughly 1h+ to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Verify against the repo before relying on details.