Own your intelligence. Keep your enterprise content governed. Using the new Huggingface Llama app, I show how to build a fully local AI agent that turns a customer RFP into a review-ready, source-cited draft, without sending a single token to a hosted API. The agent retrieves a received RFP and four approved internal source documents from Box, drafts answers to the RFP questions with named source citations, flags gaps and uncertainties instead of hallucinating, and stops for human approval before writing the final draft back into a Box Review folder. Why this matters: No API keys, no per-token cost, no data leaving your machine by default Choose the best local model (and quantization) for each workflow — swap models without moving your content Keep RFPs, approved sources, and generated drafts in one governed content layer Preserve a clear, auditable path from source material to draft output TRY BUILDING YOURSELF! Sign up for free Box developer account: https://account.box.com/signup/developer Install Llama app: https://llama.app/ Demo repo: https://github.com/box-community/box-rfp-response-demo 📚 DEVELOPER GUIDES & DOCS Everything you need to integrate Box into your agents and apps: Connect Agent to Box Guide: https://developer.box.com/tutorials/connect-an-agent-to-box Getting Started Guide: https://developer.box.com/guides/getting-started/ Box CLI Guide: https://developer.box.com/guides/cli/ API Reference: https://developer.box.com/reference/ TIMESTAMPS: 00:22 - Introducing the demo of the video 00:48 - RFP, company sources stored on Box 02:00 - Why local AI is great for confidential doc work 02:28 - How to setup local AI with Llama app 04:50 - Walkthrough AGENTS.md and script 08:07 - Running the workflow 11:12 - Conclusion 🔔 FOLLOW US ON X • Box: https://x.com/Box • Chris Kim - DevRel : https://x.com/chriskim_dev

 115      1