Build intelligent experiences: Deep dive into AppFunctions
summary
Shows how to expose app capabilities as tools via the AppFunctions Jetpack API, register them with Android’s registry, and let privileged agents call them in the Android MCP model; demoed a Jetpack travel app that uses on‑device Gemini Nano for receipt parsing (extracts title, amount, category successfully), uses the ML JIT JNI prompt API for image+prompt calls, and supports ADB/IDE testing plus a skill for end‑to‑end tests; Nano runs on 140M+ devices and Nano 4 arrives on Pixel 10; API is alpha/experimental (alpha10 changes).
AI-written summary · about
chapters (11)
- 0:00Introduction
- 0:54AI on Android
- 2:18Implement in-app intelligence
- 6:45Integrate with the intelligence system
- 10:08Task automation
- 12:18AppFunctions
- 14:47Android MCP powered by AppFunctions & Demo
- 19:25Deep dive into AppFunctions
- 23:12New features in alpha 10 & Experimental status
- 25:05Testing AppFunctions
- 27:14Experiment with AppFunctions today
related talks
28mBuild intelligent experiences: Deep dive into AppFunctions
36mElevating OS Experiences with AI-Driven Velocity
24mAppFunctions: Making Your Android App Agent-Ready
41mBuild intelligent Android apps with Google's AI
21mAdd Generative AI capabilities to your Android app
41mAgent skills for Android app development
41mGen AI for Android Developers
30mGenAI on Android with Gemini Nano and Firebase