LangChain Deep Agents + Box AI: Multi-Agent Document Analysis in a Sandbox Description A walkthrough of an end-to-end AI document pipeline built with Box AI and LangChain Deep Agents. Box AI Extract Structured pulls typed, structured data from vendor PDFs and DOCXs — no custom parser required. A sandboxed multi-agent setup (security, compliance, and contract risk subagents running in parallel) reasons over that data using an in-memory virtual filesystem. Results get written back to Box as a Markdown report with metadata on every source file, a shared link, and a review task. All the reasoning logic lives in prompts. The Box API handles the rest. What's covered: - Box AI Extract Structured for document intelligence - LangChain Deep Agents + VFS sandbox for multi-agent orchestration - Publishing results back to Box (report, metadata, shared link, review task) Links: - GitHub repo: https://github.com/box-community/langchain-deepagents-sandbox-example - Free Box developer account: https://account.box.com/signup/developer - LangChain Deep Agents docs: https://docs.langchain.com/oss/python/deepagents/overview

 98      5