Related: Google’s Gemma 4 passes 200 million downloads in just over two months.
Cursor, the AI coding tool from Anysphere, has released a native iOS app that lets developers manage their AI coding agents from a phone. The app does two things that matter, it can launch always-on cloud agents that keep working from anywhere, and it can remotely control agents already running on the user’s computer. Together they push Cursor’s pitch past the desktop editor and toward a model where the human starts and supervises the work while the software does the typing, whether or not anyone is at a keyboard.
The core idea is that coding agents no longer need a person watching them every second. Cursor’s cloud agents run in isolated virtual machines with full development environments, which means they can write code, run it, test it, and check their own output rather than just suggesting edits. Because they run asynchronously with their own tools, they can keep grinding on a long task and iterate toward a pull request that is ready to merge. The phone becomes the place where a developer kicks that process off and checks in on it, not where the work itself happens.
The app is built around that supervision loop. A user can launch several agents in parallel, track long running tasks, and inspect logs or media output straight from the phone. They can review the diffs an agent produced, merge a pull request, and annotate a screenshot to point an agent at exactly what needs fixing. Voice dictation lets someone describe an idea out loud and set a background agent working on it, which fits the on the go use case the app is chasing. The friction of opening a laptop, finding the right branch, and reading through changes is replaced by a few taps and a spoken instruction.
To make that practical, Cursor leans on native iOS features. The app uses Live Activities and push notifications so a developer gets pinged when an agent needs a decision, finishes a task, or has a pull request waiting for review. That notification model is what makes always-on agents usable from a phone, because the alternative, opening the app every few minutes to check, would defeat the purpose. The agent works quietly in the background and reaches out only when it is stuck or done, which is roughly how a person would delegate work to a capable junior colleague.
There are limits worth keeping in view. The app is in public beta and tied to paid Cursor plans, so it is aimed at existing customers rather than a general audience, and it requires a recent version of iOS. More to the point, autonomous agents still produce code that a human needs to read before it ships, which is why the app puts diff review and pull request approval front and center rather than promising hands off automation. The agent can carry a task a long way on its own, but the developer remains the one who decides what actually gets merged.
The release fits a clear direction across the industry, where the unit of work is shifting from a single AI completion to a longer running agent that owns a whole task. Cursor has been one of the fastest growing tools in that space, and putting agent control on a phone is a bet that developers want to manage this work the way they manage everything else, from wherever they happen to be. It also nudges expectations about what software development looks like, with the engineer acting more like a manager of several agents than a person writing every line. Whether that model holds up depends on how reliable the agents prove on real, messy codebases, but the iOS app makes the workflow concrete. The desk is no longer the only place the work gets done.
Related on Entrelligence: Google’s Nano Banana 2 Lite and Omni Flash models, Google’s Gemma 4 passing 200 million downloads, and Claude running on NVIDIA GPUs in Azure.
Your First 10 AI Skills
10 practical AI skills, copy-paste prompts and a 7-day plan to start using AI with confidence.
Download the guide →