AI coding agents are moving beyond autocomplete. Modern agents can run commands, call APIs, read files, and execute multi-step workflows directly from the IDE. With the Model Context Protocol (MCP), those agents can also reach beyond the local filesystem into external platforms like Box. This video shows a lightweight Cursor Plugin for Box that gives AI coding agents access to enterprise content while encoding best practices and governance guardrails. Instead of opening files, downloading documents, or copying text between tools, an agent can search Box, retrieve documents, run AI extraction, and return structured results directly inside the editor. The plugin itself is intentionally simple. It consists of just three files: • mcp.json – connects Cursor to the Box MCP server • SKILL.md – teaches the agent best practices for working with Box tools • box.mdc – enforces mandatory safety guardrails Once connected, the agent gains access to Box capabilities like file search, folder listing, metadata updates, and Box AI extraction. This makes workflows possible such as: • Finding the latest contract in a folder • Extracting renewal terms with Box AI • Updating metadata or routing files• Running multi-step document workflows The skill file guides the agent toward reliable patterns—like resolving folder IDs before searching and preferring metadata search when queries match known fields—so results are more precise and consistent. The rules file adds enterprise guardrails, requiring confirmation before destructive actions and encouraging AI processing to happen inside Box when sensitive content is involved. The result is a simple but powerful pattern: connect an AI coding agent to a governed content system and shape how it behaves using lightweight configuration and natural-language guidance. Resources Box Cursor Plugin (GitHub)https://github.com/box-community/cursor-box-plugin Create a Free Box Developer Accounthttps://account.box.com/signup/developer

 99746      16