Match an image, then click
Find a specified button, icon, or image on screen and click it automatically. The Flow follows the matched target even when its position changes.
Android automation with AI model support
Automatisation Android compatible avec les modèles d’IA
Create Android automation Flows without programming by dragging nodes. Integrate vision, OCR, LLM, and other AI models, or use AI Planner and MCP to configure Flows faster.
No programming requiredDrag-and-drop nodesAI model supportAI Planner and MCPReusable Flow Profiles
Flux de travail
Clear Android automation use cases
Read and analyze screen text
Read page messages, order status, form content, or search results, then use the text to decide, classify, or continue to the next step.
Repeat fixed actions
Connect actions such as opening an app, tapping, typing, swiping, and waiting into a Flow, then save it and run the same steps whenever needed.
Android app automation testing
Run UI steps automatically, verify buttons, text, and results, then keep screenshots and run logs for app testing and regression testing.
Automation comparison
How LaiCai Flow differs from common Android automation approaches
| Comparison | LaiCai Flow | One-time AI Agent | Recorded macros or fixed scripts | General RPA or test frameworks |
|---|---|---|---|---|
| How it is created | Drag and connect nodes. AI Planner or MCP can help configure the Flow. | Describe a temporary task for the Agent to attempt. | Record actions or write a fixed script. | Set up and maintain a larger automation project. |
| AI capabilities | Add vision, OCR, LLM, and other AI models as needed. | The Agent makes decisions while attempting the task. | Usually follows predefined actions and conditions. | Depends on the framework and custom integration. |
| Reuse and changes | Save as a reusable Profile; adjust nodes, parameters, inputs, and outputs. | Usually intended for a one-time attempt. | Page or coordinate changes may require rerecording or script edits. | Reusable, but typically requires more engineering and maintenance. |
| Best suited for | Stable Android automation that can be configured without programming. | Temporary exploration or one-off operations. | Simple repetitive actions that rarely change. | Developer-led automation and testing projects. |
FAQ
LaiCai Flow FAQ
What is LaiCai Flow?
LaiCai Flow is an Android automation tool with AI model support. Create Flows without programming by dragging and connecting nodes, integrate vision, OCR, LLM, and other AI models, or configure Flows faster with AI Planner and MCP.
Can it run independently on an Android phone without a computer connection?
LaiCai Flow runs on the computer: it reads the mirrored screen, performs image analysis and Flow logic, then sends actions to the Android phone. For independent on-phone operation, use LaiCai Flow Inside. Deploy the same compatible Flow Profile to the LaiCai Android Agent so recognition, decisions, and actions run locally on the phone without USB or a Wi-Fi connection to the computer.
View the LaiCai Flow Inside guideCan LaiCai Flow use YOLO or object detection?
Yes. The vision.detect node can run a selected local YOLO/ONNX object detection model, then output the best rectangle, center point, confidence score, class name, and detection list for later tap or branch steps.
Does LaiCai Flow support LLM text and image analysis?
Yes. llm.textAnalyze can analyze OCR, API, file, or other node output text. llm.imageAnalyze can inspect screenshots or image artifacts with a vision-capable runtime LLM and return stable structured JSON fields for later branches.
Can it recognize text, buttons, or images on the screen?
Yes. LaiCai Flow supports vision.ocr for ROI text recognition, vision.match for template image matching, ui.find for Android UIAutomator element search, and vision.capture for screenshots used by LLM or downstream nodes.
What Android actions can a Flow run?
A Flow can tap, swipe, long press, drag, type text, send keys, set or read clipboard text, open or force-stop apps, go Home or Back, then organize the work with wait, branch, repeat, for-each, until, and child Flow nodes.
Can a Flow speak or play voice prompts?
Yes, with one boundary: audio.speak plays text aloud on the desktop LaiCai computer using desktop text-to-speech from the local speech engine. It is for spoken reminders or OCR/API result announcements, not Android device-side TTS.
Can Flow connect HTTP, ADB, AutoJS, or macro steps?
Yes. LaiCai Flow includes http.request for external APIs, adb.shell for Android shell commands, autojs.run for existing AutoJS scripts, macro.play for saved macros, plus file, template image, and storage data nodes.