AppClaw CLI
A terminal agent for phones: state a goal in plain words and it reads the screen, then decides where to tap and what to type
- 120 ★ on the repository
- Last commit
Broad access
Why this level
- The agent taps and types in real apps on the device and may send a message or change settings.
- Screen content is sent to the model provider.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/appclawDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The agent taps and types in real apps on the device and may send a message or change settings.
- Screen content is sent to the model provider.
This is third-party code. Review the repository files before installing.
What it does
AppClaw runs an AI agent that drives an Android device, an emulator or an iOS simulator from a plain-language goal. The agent reads the screen, picks a tap, text entry or swipe and repeats until the goal is done. Besides agent mode there are deterministic YAML flows with no model calls, where selectors and state assertions are explicit. `appclaw test` runs vitest-style specs on real devices and `appclaw init` scaffolds a project. Terminal Studio shows the device screen next to the agent's progress.
Who it is for. Mobile QA engineers and app developers who want to run a scenario on a phone without hand-writing selectors.
Good fit when
- A check is described in words and no selectors exist yet.
- The same run is needed across several devices and OS versions.
- A terminal agent has to open an app on an emulator and reach a given screen.
- A successful run should become a repeatable YAML flow.
Not a fit when
- The app is already covered by stable tests with ready selectors.
- A device with personal accounts must not be left to an unattended agent.
- Screen content must not be sent to an external model provider.
Example request
Open settings on the Android emulator and turn on airplane mode, then save the run as a YAML flow.Limitations
Node.js 22 or newer, a connected device, emulator or simulator and a model key from Anthropic, OpenAI, Google or Groq are required; local Ollama needs no key. Screen content goes to the chosen model provider unless a local model is used. The agent acts in real apps, so test on a separate account. `appclaw doctor` checks the environment and prints fix hints; with `--full` it also starts appium-mcp.
How to disable. Remove the package with `npm uninstall -g @appclaw/cli` and delete the model key from the project `.env` file.
FAQ
Which devices does AppClaw support?
Android devices and emulators, plus iOS simulators. The screen mirror in Terminal Studio works with the same set.
How do I run a scenario without model calls?
Describe the steps in a YAML file with selectors and state assertions. Such a run is deterministic and spends no tokens.
Which models can I connect?
Anthropic, OpenAI, Google, Groq and local Ollama. The key goes into the `.env` file.
How do I check the environment is ready?
Run `appclaw doctor`: it checks Node.js, `.env`, the model key, appium-mcp and devices and prints fix hints.
What is Terminal Studio?
A full-screen shell opened by the `appclaw` command. You give goals one after another and see a device mirror; `^r` starts the mirror and `^x` closes it.
Discussion
No questions or reviews yet.
Related
CLI toolsSkill and CLI for controlling an Android phone via ADB and vision models: the agent opens apps, searches, orders and verifies results
mobilecli drives iOS and Android devices, emulators and apps from the terminal, and its skill teaches Claude Code how to use it
62 MCP tools so an agent can tap, scroll and type on a real Android phone over ADB or on an iPhone.
Lets Claude Code and Codex drive a real iPhone: read the screen, tap, type and scroll in any app
Skills and a CLI that let Claude Code, Codex and OpenCode operate an Android phone: open apps, tap, type and test workflows
An MCP server to drive iOS and Android: simulators, emulators and real devices via accessibility trees and screenshots
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| AppClaw (this card) | no | 120 | ||
| Mobile MCP | no | not required | 10,325 | exposes phone control to an external agent over MCP; AppClaw drives the goal itself and writes YAML flows |
| Appium MCP | yes | not required | 518 | provides Appium tools; AppClaw adds an agent loop, a test runner and Terminal Studio |
README in short
Installation goes through npm and `appclaw doctor` checks the environment. Packages are split into the CLI, a test runner and an SDK for embedding in vitest, jest or mocha.