Mika
An on-device Android AI agent that operates phone apps and the web on a hidden background screen.
Mika executes multi-step app operations, web research, and routines locally on your Android device.
Mika is an open-source on-device AI agent for Android that interacts with native applications and web pages using computer vision and UI automation. It runs tasks on an isolated background screen so your main display remains usable while it works. This tool suits power users and executives who want to delegate tedious mobile app actions without routing data through third-party servers.
What it is
Mika reads Android screens like a human user, typing, scrolling, and tapping elements across third-party apps without needing dedicated APIs. It features a built-in browser for web tasks, an isolated Linux sandbox (arm64) for running Python and shell scripts, and scheduling for automated routines. You can monitor progress through a live picture-in-picture view or intervene whenever safety checks or payment prompts arise.
Who it's for
Mika suits Android power users, mobile automation enthusiasts, and executives who want an on-device personal assistant that works directly inside mobile apps. It does not suit iOS users, as it requires Android 11+ (14+ recommended) and wireless debugging setup. It is also less ideal if you prefer a zero-configuration solution that requires no model API key setup.
Key features
Background screen execution
Runs tasks on an isolated virtual screen, allowing you to use your phone normally while watching via a live preview thumbnail.
Universal app and web automation
Navigates app interfaces and websites visually by identifying buttons and fields without needing per-app integrations or official APIs.
Built-in Linux sandbox
Executes Python, Node, and shell scripts on local files directly within a per-conversation container on arm64 devices.
Computer CLI and MCP integration
Allows AI tools running on your desktop computer to hand off tasks to Mika using a command-line interface or Model Context Protocol.
On-device memory and skill creation
Remembers preferences and automatically converts completed workflows into reusable skills to speed up future executions.
Up and running in 3 steps
Our honest verdict
What's great
- Operates apps on an isolated background screen without blocking your active device display.
- Stores API keys, memories, and task logs strictly on-device without a central telemetry server.
- Includes an arm64 Linux sandbox to execute Python, Node, and shell scripts directly on mobile files.
- Integrates with desktop tools via CLI and Model Context Protocol (MCP) for cross-device workflows.
Where it falls short
- Requires wireless debugging pairing via Android settings, which can turn off after device reboots or Wi-Fi changes.
- App and website automation depends on visual recognition, which may break when app layouts change.
- Requires your own LLM API key (OpenAI, Anthropic, DeepSeek, etc.) or account login to function, creating ongoing API usage costs.
Pricing & license
Mika is completely free and open-source under the Apache-2.0 license. Because there is no hosted service provided by the author, your only cost is the API usage fees billed directly by your chosen LLM provider (or zero cost if using a local server like Ollama).
Open remote AI-training roles
873 open remote AI-training roles, updated November 15, 2026; most pay $20–$170 an hour. Zekai lists them free; the application goes to the employer’s platform.
Alternatives
These reviewed AI tools do the same job instantly — from our scored directory.
Use it, or have us set it up
FAQ
Does Mika run a telemetry or backend server?
No. There is no Mika backend server; your input and screen captures are sent directly to the model provider you configure.
How does Mika operate apps without blocking my phone?
It launches and operates apps inside a separate virtual background screen while giving you a small live view window to observe or take over.
What model providers and accounts can I connect?
Mika includes presets for providers like OpenAI, Anthropic, Gemini, DeepSeek, Ollama, and OpenRouter, or you can supply any OpenAI-compatible API endpoint.
What happens during payments or password screens?
Security-sensitive screens like payment checkouts or passcode inputs block screen captures, causing Mika to pause and hand control back to you.
