PhoneUse Launches Dedicated Cloud-Based Real Devices for AI Agents
Abdulla Ababakre provides dedicated cloud-based real Android and iPhone devices for AI agents through the PhoneUse platform, allowing agents to perform operations directly on the devices without relying on application open interfaces.
This solution addresses three major limitations: most mobile applications (including Uber, Instacart, TikTok, Instagram, WhatsApp, Telegram, and many traditional apps) lack APIs or MCPs, or strictly limit automated access, and browser proxies cannot access native application environments; platforms like Amazon block shopping proxies, airlines restrict automatic bookings, and financial institutions limit proxy operations; proxies cannot read mobile notifications, SMS, or real-time application statuses, and cannot proactively respond to changes.
The platform currently has 312 devices online, of which 288 are cloud Android devices with a startup time of about 4 seconds; 24 are real iPhone, Pixel, and Galaxy devices deployed in a San Francisco rack, using non-jailbroken retail models, real carrier SIM cards, physical stylus input, and top cameras to read screens, passing device integrity checks.
On-demand cloud phone billing is $0.024 per minute, with a real iPhone monthly fee of $89; the same MCP interface or API can switch between cloud and real devices, supporting connections to clients like Claude, Cursor, Codex, and Windsurf, and can also be called via CLI without local device configuration.
Demonstration scenarios include completing a Diet Coke purchase on Amazon, posting or interacting on TikTok and Instagram, processing Geico and Reddit application workflows, and proactively sending SMS notifications to contacts after receiving flight delay push notifications.
AI agent developers and enterprises purchase cloud or real device computing power to bypass closed application ecosystems and anti-automation detection, driving event-driven funding towards mobile automation tools; beneficiaries include agent applications needing native application interaction and mobile-first services, while platforms relying on browser or API restrictions to control automation face pressure.
The platform plans to scale device provision in data centers and will soon open-source local mobile operation algorithm code.
Source: Public Information
ABAB AI Insight
Abdulla Ababakre previously served as CEO of InterviewCoder, a product positioned as a technical interview assistance tool, emphasizing screen sharing and undetectability in monitored environments, having reported serving over 90,000 users and generating millions in annual recurring revenue; he also operates PhoneFarm, which faced scrutiny after allowing agents to control real phones, resulting in his personal account being suspended, after which he publicly clarified to the platform and emphasized its use for anti-bot detection.
The current path is to transfer the covert operation and device interaction experience accumulated from InterviewCoder to provide independent mobile resources for agents; achieving low-cost scaling through cloud virtual devices while retaining physical real device racks to handle high detection threshold applications, forming a resource reallocation from interview tool revenue to agent infrastructure, motivated by the need to capture the window of agents expanding from browsers to mobile.
This is similar to the evolution path of computer-use and subsequent browser-use tools promoted by OpenAI and Anthropic, and is akin to early cloud gaming or device farms (such as real device rack services used for application testing); currently in the early stage of agent capabilities expanding from desktop and web to mobile native environments, the platform aims to establish a device supply advantage in detection countermeasures.
Essentially, this is a technological substitution: by providing real devices and a physical input layer, it bypasses the closed application layer APIs and integrity detection mechanisms; the reason is that most mobile applications deliberately do not open automation interfaces to protect their ecosystems, while agents need to perform human daily operations, making physical devices the only currently detectable universal entry point, thus shifting control from application providers to device providers.
ABAB News · Cognitive Laws
Where interfaces are closed, physical devices are the passport
Agent capability = operable devices × detection pass rate
From browser to mobile is a shift from web permissions to life permissions