Analysis updated 2026-05-18
Set up a phone automation using a plain-language description instead of building a rule by hand
Automatically send yourself information or toggle phone settings based on time, location, or events
Review and approve exactly what an automation rule will do before it can ever run
| jackrushante/argus | androosio/thortune | azcomp2000/battery-sentinel | |
|---|---|---|---|
| Stars | 0 | 0 | 0 |
| Language | Kotlin | Kotlin | Kotlin |
| Setup difficulty | moderate | easy | moderate |
| Complexity | 4/5 | 2/5 | 2/5 |
| Audience | developer | general | developer |
Figures from each repo's GitHub metadata at analysis time.
Runs as a normal Android app, elevated features need the optional Shizuku tool and Android 14 or newer.
Argus is an Android automation app similar in spirit to Tasker, but built around a specific idea: an AI language model only writes the rule, it never runs the rule. You describe what you want in plain language, for example sending you the Bitcoin price every day at nine, or turning off Wi-Fi when you leave home. The language model turns that description into a structured set of triggers, conditions, and actions using a fixed vocabulary of types. You then review and approve the exact rule, shown from its underlying data rather than from the AIs own wording, and it gets locked in with a cryptographic fingerprint. From that point on, a separate deterministic engine runs the rule using the phones own system features, such as alarms, geofences, and notification listeners, without consulting the language model again. The only exception is an optional action that deliberately calls the model again at the moment it fires, and even that runs inside a tightly bounded contract you already approved, with limits on tools, timeout, and budget. The project explains this split matters for three reasons: the rule that fires is always exactly the one you approved, you are not paying for or waiting on a model call every single time a rule triggers, and untrusted content like an incoming text message can at most flip a switch you already agreed to, never invent a new command on its own. Argus supports several language model providers, including a bridge to a local command line tool as well as OpenAI-compatible, Gemini, OpenRouter, and Anthropic APIs. Elevated device permissions are optional through the Shizuku tool, and a basic tier of the app works without them. The app has no backend server, no account, and no telemetry: rules, logs, and API keys stay on the device, with keys stored encrypted. Argus is written entirely in Kotlin, targets Android 14 and newer, has a bilingual English and Italian interface, and is released under the GPL-3.0 license, meaning anyone can study, modify, and share it, but any version built on it must stay open under the same terms.
An Android automation app where an AI turns your plain-language rules into structured triggers and actions, which then run on a separate, predictable engine with no AI involved at run time.
Mainly Kotlin. The stack also includes Kotlin, Android, Jetpack Compose.
GPL-3.0: you can study, modify, and share the app freely, but any modified version must also stay open source under the same license.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Verify against the repo before relying on details.