Skip to main content

Documentation Index

Fetch the complete documentation index at: https://developer.box.com/llms.txt

Use this file to discover all available pages before exploring further.

Use the Box MCP server with LangChain Agent Builder to build reusable agents that securely connect to Box, analyze files with Box AI, extract structured data, and save artifacts back to Box.

​
Prerequisites

  • A Box enterprise account with the MCP server enabled by an admin. See Set up the MCP server.
  • A LangSmith account with access to Agent Builder.

​
Enable Box integration in Box

Admin
  1. Sign in to the Box Admin Console.
  2. Go to Integrations and find Custom Box MCP Server either by using the MCP Category filter, or by using the search bar at the top of the page.
  3. Go to Configuration > Add Integration Credentials to generate new credentials.
  4. In Redirect URIs, change the Box redirect URIs to the Redirect URI provided by the external MCP Client.
  5. Copy the Client ID and Client Secret for later use. These are required later for LangSmith to authorize the connection.
  6. Check the Access scopes:
    • Scopes define the maximum set of actions. Users can only access content they already have permission to view or edit in Box.
    • Read all files and folders stored in Box.
    • Read and write all files and folders stored in Box.
    • Manage AI Requests.
  7. Click Save.

​
Configure Connection in LangSmith Agent Builder

Admin Follow the instructions in the LangChain’s documentation: Manage workspace administration, Remote MCP Servers.

​
Video walkthrough

The video below shows how to use the Box MCP server with LangChain Agent Builder to build a reusable intake review agent. The agent securely connects to Box, analyzes files with Box AI, extracts structured data, and saves a review packet back to the source folder. In this guide, you will:
  • Connect LangChain Agent Builder to Box using MCP and OAuth
  • Use a prebuilt agent template for intake review workflows
  • Run the agent with a simple prompt
  • Extract key fields, classify documents, and flag missing information
  • Generate a durable review artifact saved back into Box