droidrun/mobilerun is an LLM agent framework that automatically operates Android and iOS devices with natural-language commands. The latest public release is v0.7.0, and neither the change list included in this version nor its release notes are present in the given record.
The design of this repository is not tied to a specific language model. LLM-agnostic means that any language model can be connected and used. The README explains that it reads the device's UI state and screenshots, then plans actions such as tap, swipe, and input in order, and carries out app tasks broken into multiple steps. The results are exposed through a CLI and a Python API. The CLI is the channel used from the command line, and the Python API is the way to call it as a function from other programs.
The implementation language is Python and the license is MIT. The repository description is summarized in one line: automating devices with natural-language commands. The repository was created on April 12, 2025 and has 9,712 GitHub stars.
That said, there is nothing here to gauge the accuracy of its device operation. No success rate or per-task benchmark figures are given, and even which model is connected in what way is not recorded. Of the score items, only momentum is filled in, at 4.7 out of 10, while the freshness, technology, and quality items are empty. Since only the MIT license, the Python implementation language, and the feature scope documented in the README can be confirmed, a performance-based adoption decision would be premature. If you are planning to automate phone operation, the sensible course is to check the release notes and performance data further, and then treat it as a candidate for review.