A Smarter Google AI Edge Gallery: MCP integration, notifications, and session continuity

2026-05-19 · Source: Google Developers Blog - AI · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Intermediate, medium

Summary

Google AI Edge Gallery, an on-device AI showcase app for Android and iOS, has been updated to support the Model Context Protocol (MCP), local notification reminders, and persistent chat history. This expansion enables developers to build more connected, proactive, and persistent agentic experiences using Gemma and other open models directly on mobile devices. The MCP integration, currently an experimental feature on Android, allows on-device LLMs to interact with external tools and data sources via Streamable HTTP, with reasoning and decision-making occurring locally. New "Schedule Notification" skills facilitate automated routines, such as mood tracking or calendar briefings. Additionally, the app now offers persistent chat history, leveraging LiteRT-LM backend's fast prefill capability (exceeding 3,000 tokens per second), and allows direct editing of custom system prompts for enhanced developer control.

Key takeaway

For AI Architects and Machine Learning Engineers building mobile-first AI applications, the Google AI Edge Gallery updates offer critical capabilities for creating sophisticated on-device agents. You should explore integrating the Model Context Protocol (MCP) to enable your LLMs to interact with external tools and data sources, and leverage the new notification and persistent chat features to build proactive, stateful user experiences. Consider the open-source toolkit and community contributions for accelerating your development of tailored, utility-focused workflows.

Key insights

Google AI Edge Gallery enhances on-device AI with MCP, notifications, and persistent chat for agentic mobile experiences.

Principles

• On-device LLMs require standardized external interaction.
• Smaller context windows necessitate concise tool descriptions.
• Proactive interactions enhance automated use cases.

Method

Integrate MCP via a registered URL to dynamically import tool definitions and resource schemas into the on-device model's system prompt, enabling local reasoning and tool execution by an MCP server.

In practice

• Use MCP for on-device LLM tool calling.
• Keep MCP tool descriptions short for efficiency.
• Experiment with custom system prompts for agent personas.

Topics

• Google AI Edge Gallery
• Model Context Protocol
• On-device AI
• Agentic Workflows
• Local Notifications

Code references

• google-ai-edge/gallery

Best for: AI Architect, Machine Learning Engineer, NLP Engineer, AI Engineer, Software Engineer, Prompt Engineer

Related on AIssential

• Google AI Edge Gallery — Google's new app enables on-device execution of Gemma models and tool calling directly on iPhones.
• Building agents that reach production systems with MCP — MCP standardizes agent-to-system integration, enabling scalable, portable, and context-efficient connections for cloud-based AI agents.
• Google Pushes for gRPC Support in Model Context Protocol — Google's gRPC transport for MCP aims to bridge AI agent protocols with existing microservice architectures.
• The MCP: New brain of artificial intelligence — The Model Context Protocol (MCP) standardizes AI interoperability, transforming isolated models into connected, orchestrating systems.
• New ChatGPT Library Explained & More AI News You Can Use — AI platforms are evolving to support document-centric workflows and simplified multiplayer application development.
• Making AI Personal: Building Context-Rich AI Tools with MCP • Kris Jenkins • GOTO 2025 — Agentic AI combines LLMs with programmer-defined tools, enabling context-rich answers to ambiguous personal queries.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Google Developers Blog - AI.