Today's AI open source · 2026-11-11
wangzan101/Mika
Mika is an AI agent app that operates Android apps and the web for you from a single spoken or typed request.
| Repository | github.com/wangzan101/Mika ↗ |
|---|---|
| Main language | Kotlin |
| License | Apache-2.0 |
| GitHub stars · forks | ★ 8,070 · 1,840 forks (as of 2026-11-11) |
| Created · last update | 2026-10-29 · 2026-11-10 |
| GitHub topics | accessibility, ai-agent, android, automation, jetpack-compose, kotlin, llm, mobile-agent |
Overview
Mika is an open-source AI agent app that runs on Android phones. Describe what you want done in one sentence, and it opens apps and websites itself, tapping, typing and scrolling until the task is finished.
Tasks run on a separate, invisible background screen, so you can keep using your phone normally in the meantime. Without app APIs or per-app integrations, it looks at the screen and finds buttons the way a person would, and in the built-in browser you can give instructions directly on the web page you are viewing. There is no separate Mika server; the phone connects directly to the model API you choose.
It suits users who want to hand off repetitive tasks that span several apps, such as summarizing messenger chats, shopping or comparing flights, to their phone. It is also a useful reference for developers who want to study how mobile agents work through code.
Why it is worth a look
What stands out is that it works by looking at and operating the screen rather than building integrations for each app, so it can be applied to any app. The design in which the agent works on a separate background screen, without taking over the user's screen, directly targets a common inconvenience of phone agents. The README also describes turning completed tasks into Skills so they can be handled faster next time.
Key features
- Opens apps on a separate background screen and taps, types and scrolls, with progress visible in a small live view.
- Finds buttons by looking at the screen without app APIs, so it is not tied to specific apps.
- Lets you give instructions directly on the page you are viewing in the built-in browser, and stops when you touch the page.
- Remembers preferences you tell it, and turns completed tasks into Skills for use in later runs.
- Asks for confirmation before orders and payments, and flags logins and verification codes for the user to handle.
What it is used for
- Use it to read and summarize today's conversation in a messenger group chat.
- Use it to pick products that meet your criteria, add them to the cart and stop just before payment.
- Use it to compare flights that meet your criteria on an airline ticket site.
- Use it to read through a specific creator's recent posts one by one and summarize them.
Getting started
- Download the APK from GitHub Releases and install it (Android 11 or later, arm64).
- In Settings › Model, choose a provider and enter an API key, or sign in with a ChatGPT, xAI or OpenRouter account.
- To control phone apps, follow the 'Phone control setup' wizard in Settings to pair wireless debugging once.
- Describe what you want done in one sentence and run it.
What to weigh
- Released under the Apache-2.0 license.
- It is an early-stage project created in late October 2026, so features and settings may change.
- A model API key or an external account is required, and model costs apply based on usage.
- Because it grants broad phone control permissions through wireless debugging, you should consider in advance which apps and tasks to entrust to it, and each app's terms on automated operation.
Related terms and companies
Sources
Written with AI assistance from the repository README and GitHub data. Star counts and other numbers are GitHub values as of the date shown; features and licensing can change, so check the repository.