Microsoft Foundry Local + CAG: Build Fully Offline AI Apps with Two NPM Dependencies and Zero Cloud

Microsoft Community Hub·high signal

Microsoft published a hands-on tutorial for building a fully offline AI support agent using Foundry Local and Context-Augmented Generation (CAG). Unlike RAG, CAG pre-loads your entire knowledge base into the model context at startup—no vector DB, no embeddings, no chunking. The total dependency footprint is two npm packages (express + foundry-local-sdk). The system runs entirely on-device with four layers: browser client, Express server, CAG engine, and Foundry Local AI layer on CPU/NPU.