Mobai MCP Server

Android and iOS automation in mobai-mcp starts with discovering connected devices, emulators, and simulators and checking their details. Automation requires the local MobAI desktop app and an active on-device bridge, which the tools can start or stop. For native apps, an assistant can list installed applications, launch one by bundle ID, inspect its accessibility tree, and tap indexed elements or screen coordinates. Text entry, swipes, and returning to the home screen are also supported. Screenshot capture returns a saved PNG path, while iOS-only OCR supplies recognized text and coordinates for tapping.

Browser and WebView workflows include finding pages, navigating, inspecting the DOM, clicking or entering text through CSS selectors, and executing JavaScript. The execute_dsl tool groups automation actions into batches, supporting observation, scrolling, waits, assertions, and conditional checks. Alternatively, run_agent accepts a task for an agent to pursue by observing the screen, choosing actions, and performing them. For operations outside the named tools, http_request provides direct access to the MobAI API. MCP resources supply API documentation and guides for DSL batches, native apps, and web automation. Clients communicate with the server through stdio.