AppClaw logo
Open Source · Apache 2.0

The intelligent agent for mobile

Describe what you want in plain English. AppClaw figures out what to tap, type, and swipe — on real Android and iOS devices.

Get Started → View on GitHub
appclaw — zsh
$ npx appclaw

? Select mode: Agent
? Select LLM: claude (sonnet)
? Goal: Open WhatsApp and send "Hello!" to Mom

● Connected to device emulator-5554
● Reading screen...
Step 1: Tap on WhatsApp icon
Step 2: Tap on search bar, type "Mom"
Step 3: Tap on "Mom" conversation
Step 4: Tap message field, type "Hello!"
Step 5: Tap send button

✔ Goal completed — 5 actions, 12.4s
AppClaw automating a mobile device in real time
Works with any LLM provider
A
Anthropic
O
OpenAI
G
Google
G
Groq
O
Ollama

Three ways to automate

AI agent for complex goals, YAML flows for repeatable tests, or an interactive playground to build live.

Agent Mode

From prompt to completed action

Describe a goal in plain English. AppClaw reads the screen, reasons, and executes until done.

Agent Mode
$ appclaw "Turn on WiFi"

1/3 launch → Settings
2/3 tap → Network & internet
3/3 tap → Wi-Fi toggle

✔ Done — 3 steps, 8s, $0.002
YAML Flows

Deterministic steps, zero LLM cost

Write plain English YAML. No AI needed. Fast, repeatable, with vision fallback.

login-flow.yaml
steps:
  - open YouTube app
  - click on search icon
  - type "Appium 3.0"
  - perform search
  - scroll down until "TestMu" visible
done:
  - "Search results visible"
Playground

Build flows interactively

Live REPL on a real device. Type commands, watch them run, export to YAML.

Playground
pg> tap on Settings
✔ Tapped "Settings"

pg> scroll to General
✔ "General" visible

pg> tap General
✔ Screen changed

pg> /export flow.yaml
✔ Saved (3 steps)

How it works

Three steps from intent to execution. No boilerplate, no selectors, no brittle locators.

01

Describe your goal

Write what you want to accomplish in plain English. AppClaw accepts goals as prompts, YAML files, or interactive playground commands.

$ npx appclaw
? Goal: Order a coffee from the Starbucks app
02

AI perceives and reasons

AppClaw takes a screenshot, sends it to your chosen LLM, and receives a structured plan. It understands context, handles popups, and adapts to changing screens.

Reasoning: The Starbucks app shows the home screen.
I see the "Order" tab at the bottom. I'll tap it
to begin the ordering flow.

Action: tap(element: "Order tab")
03

Execute and verify

Each action is executed via Appium through the MCP protocol. AppClaw verifies the result after every step and continues until the goal is achieved.

Step 1: Tap "Order" tab ✔
Step 2: Tap "Hot Coffees" ✔
Step 3: Tap "Caffè Latte" ✔
Step 4: Tap "Add to Order" ✔

✔ Goal completed — 4 actions, 9.8s

Pure agentic brain, zero device logic

AppClaw is an intelligent orchestrator. All device interaction happens through the open MCP protocol — no vendor lock-in, no proprietary drivers.

AppClaw
AI Agent
→
MCP Protocol
JSON-RPC
→
appium-mcp
32 Tools
→
Appium
Automation
→
Device
Android / iOS
3
Modes
32
MCP Tools
16
Actions
5
Providers
2
Platforms
AppClaw lets me automate complex mobile workflows that would take hours of manual scripting — in minutes. Just describe what you want and watch it happen.
Mobile QA Engineers
Teams using AppClaw for automated testing

Ready to automate your
mobile apps?

Open source, extensible, and ready to use. Bring your own LLM key and start automating in under a minute.

$ npx appclaw
Get Started → GitHub
Apache 2.0·BYO LLM Key·Community Driven