As AI becomes increasingly embedded in our daily lives, AI for phones, wearables, medical devices, cameras, drones will become personal. The models that use private data for context should run on-device without communicating with external servers. Cactus is a lightweight, high-performance framework for running AI models locally on mobile devices.
Cactus has unified and consistent APIs across:
- React/JS/TS for cross-platform apps
- Flutter/Dart for cross-platform apps
- Kotlin/Java for Android
- Swift/Objective-C for iOS
- C/C++ for core builds
We are redefining on-device AI with the following features:
- Model Context Protocol (MCP)
- Private phone tool use (camera control, gallery search, mail handling etc.)
- Multi-agent support and workflows on-device
- Thinking mode & on-device reasoning
- Hardware-aware models (both SOTA ports and foundation models)
We're committed to keeping our core technology open-source, fostering a community where developers can collaborate to push the boundaries of what's possible with on-device AI.
|
Experienced Product leader and a former quantitative researcher. Holds an MSc in Financial Economics from the University of Oxford, and previously worked in equity investing and algorithmic trading. A first-hand experience of the importance of data security has motivated the desire to enable on-device, private, and local AI. |
Studied Electrical and Computer Engineering, then an MSc in Artificial Intelligence from QMUL under Prof. Matthew Purver. Henry researched Maxar satellite perception models at WISDM, then edge mobile-device models at Deep Render. He contributes to and authored many open-source projects on GPU programming and Edge AI. |
The apps demonstrate how Cactus lets you:
- Run powerful AI models completely offline
- Build apps that don't need cloud connections
- Create seamless AI experiences that work anywhere
