Roger

Phonedriver

Lets AI agents see and use your phone like a human tester

About

Phonedriver is a middleware platform that allows advanced AI agents to operate real Android or iOS devices in a controlled and auditable way. By exposing device screens as structured data and enabling tap, scroll, and text input actions, it lets AI models such as Claude Code, Codex, and Cursor interact with apps just like a human would. This includes reading screen content as text, automating complex sequences, and even making purchases—always prompting a user for approval before any sensitive action, such as a transaction, is completed.

Designed to bridge the gap between digital AI agents and the physical world via smartphones, Phonedriver runs as a server and CLI that any Model Control Protocol (MCP) client can connect to. This opens up new possibilities for task automation, integrating AI into workflows that demand mobile-specific actions, such as ordering food, filling forms, or navigating apps that don’t have public APIs. The setup process uses standard developer tooling, making it accessible to engineers who want to prototype or deploy agents with real device access.

Phonedriver’s architecture emphasizes privacy and user control. Each potentially risky action—especially payments—is surfaced with a prompt, and users can inspect or modify actions before they’re executed on their device. Combined with transparent logs and compatibility with open-source agent tools, this makes Phonedriver a flexible yet safe bridge between AI and smartphones for automation enthusiasts, developers, and researchers.

Who is Phonedriver made for?

Software Developer / Engineer CTO / Head of Engineering Product Manager
Solo (1 person) Small team (2-5 people) Startup (6-10 people)

Phonedriver is meant for software developers, technical leads, and AI researchers who need to connect AI agents to real mobile devices for automation, prototyping, or testing. It’s particularly useful for teams building custom workflows that require direct app interactions on Android or iOS, including use cases where web automation is insufficient because the action must occur in-app, such as mobile order placement, in-app form filling, or navigation of app-only services.

Engineering departments in startups or small to mid-sized companies doing AI integrations, mobile process automation, or advanced QA testing will find it appealing. It’s also suitable for research labs exploring agent-environment interaction, and for product managers exploring new AI-driven mobile features who want a safe way to test ideas on real devices.

Is Phonedriver yours? Claim this listing to manage it and get featured.

Get tools like Phonedriver every Tuesday

The best new AI agent frameworks & MCP, in one short email. Free.