Your AI agent can finally touch your mobile apps

What it does

Argent is an agentic toolkit that plugs AI assistants into iOS Simulators and Android Emulators via MCP. It exposes a full control surface—taps, swipes, typing, hardware buttons, profiling, log inspection, and crash capture—so an agent can navigate apps, reproduce issues, and analyze performance exactly as a user would. React Native projects are handled natively, with builds and iterations requiring no extra setup.

The interesting bit

This goes deeper than simple emulator orchestration: the agent gets a genuine sensory and motor toolkit for mobile apps, including Xcode Instruments and React Native profiling sessions. That depth makes sense given the authors—Software Mansion are core React Native contributors, and the RN workflow feels first-class rather than bolted-on.

Key highlights

• Full UI automation for simulators: taps, gestures, pinches, and hardware buttons
• Built-in profiling support for both React Native and Xcode Instruments
• Operates locally over MCP stdio; no telemetry, analytics, or crash reporting
• Auto-detects and configures MCP for Claude Code, Cursor, VS Code, Windsurf, Zed, Gemini CLI, Codex CLI, and opencode
• Mixed licensing: source code under Apache 2.0, but simulator binaries and iOS devtools dylibs are proprietary and cannot be redistributed

Caveats

• iOS support requires macOS and Xcode; Android needs SDK Platform Tools and emulator setup
• Core simulator binaries and iOS devtools libraries are proprietary, restricting redistribution and deep inspection

Verdict

Mobile developers tired of context-switching to tap through simulators—particularly React Native teams—are the obvious audience. If your workflow never touches iOS or Android emulators, or if you require 100% auditable open-source binaries, this won’t fit.