Skip to content

Getting started

FAQ quick reference

License, LLM, Token — how do they relate?
  • License: Buy USB device license (not screen), plus IPA / BLE setup → Prerequisites & licensing
  • LLM: Only AI chat and AI workflow authoring need it; running saved flows does not → Model config
  • Token: Trial / run workflows do not consume chat Token → FAQ details

→ FAQ quick reference (full)

Before you start

  1. Open the new control center (新中控.exe — New Control Center — in the install folder)
  2. Confirm devices are connected normally (same as screen mirroring and scripts)
  3. Completed USB licensing and runtime environment in Prerequisites & licensing

What AI Agent helps you do

No coding — connect phones, speak in Chinese or click a few buttons, and run many iPhones / iPads in parallel. Good for ops, batch testing, repetitive tasks.

Speak and go (AI chat)

Tell AI in plain language; it runs on selected phones, for example:

  • “Open WeChat on every phone”
  • “Non-automation screenshot on the first one” or “Non-automation OCR — is there a login button?”
  • “What automation can I run?”

If info is missing, AI asks follow-ups (e.g. which device) instead of guessing.

Build once, reuse (workflows)

Turn fixed routines like “open App → tap here → wait 3s → type → save screenshot” into a flowchart, then:

  • One-click run from the workflow library
  • Many phones at once; progress in the task panel
  • Need manual captcha on a step? Flow pauses; finish and click Continue

Don’t know how to draw a flow? Let AI write it

In the workflow editor, describe in Chinese, e.g. “Open Safari, visit a URL, then screenshot” — AI can generate the flowchart for you to trial-run, tweak, and save.

Typical scenarios

ScenarioWhat you can do
Multi-device batch opsSelect 10, 50 devices, run one flow together
App walkthroughNon-automation/automation screenshot, OCR, template/color-offset match, VLM locate, auto tap/swipe
Login / register / checkoutSave as workflow; run on schedule or on demand
Album asset deliveryImport images/videos from PC folders to device albums
Input method scenariosDedicated IME steps: type, paste, read clipboard
Connection recoveryAdd reset USB, wait reconnect in flow to reduce manual unplug

Three-step onboarding

Step 1: Open AI Agent

  1. Start the new control center
  2. Click “AI Agent”
  3. Default tab is AI chat; if the top bar says not connected, click “Start AI”

For Chinese chat to drive devices, configure an LLM first:

  1. Toolbar “Model config”
  2. Click Add, fill in your provider — DeepSeek, OpenAI, local Ollama, etc.:
  • Name: memorable label, e.g. “DeepSeek main”
  • API URL, model name, API key: per provider docs
  • Quick templates for DeepSeek / OpenAI / Ollama — pick one and set the key
  1. Save, then select the config below the chat input

Step 3: Select devices

On the left:

  • Devices: check phones to control (multi-select)
  • Groups: pick many at once by screen mirroring group

After selection, tell AI “on these phones…”

How to know you’re ready

Where to lookOK when
Top connection statusConnected or green indicator
Model nameCurrent model visible in toolbar
Left devicesOnline devices checked
Workflow libraryCan New or see existing flows

Where data is stored (usually no change)

By default, workflows, chat history, model config live under aibin/data/aiagent/ in the install directory.

If you change path in “Settings”→ Data storage directory, content moves there; restart AI after (“Start AI” or reopen window).

Run logs always stay in aibin/log/, independent of data directory.

Next steps