Thanks to Kamil, and
Addressing Challenges of AI in Enterprise Data
Kamil opened by identifying a pain point for enterprises: the disconnect between AI tools and enterprise content. According to Kamil, this separation leads to underutilized data and inefficiencies. Recording clip here
"AI tools are disconnected from enterprise content. Employees don't get full value when the AI tools and the data they have are not connected together."
Furthermore, security risks abound when employees upload sensitive files to external AI platforms, exposing enterprises to potential data leaks, compliance violations, and governance pitfalls. Custom-built integrations, while a potential solution, pose another challenge as they are complicated, expensive, and difficult to sustain as AI technology scales and evolves.
Introducing MCP
To tackle these challenges, Box has adopted the open-source Model Context Protocol (MCP), which provides a standardized and secure way for AI systems to access enterprise data. MCP empowers AI agents to work within a governed ecosystem by creating managed interactions with enterprise content.
"With MCP, you can connect to third-party systems like Box for a single open standard... scale across multiple vendors and environments with one consistent integration model."
"MCP lets AI systems access tools and data in a consistent and secure way."
This approach eliminates manual downloading and uploading of files by employees, streamlining workflows and safeguarding sensitive data. By connecting AI agents directly to Box via the MCP server, employees gain access to real-time enterprise data while maintaining content governance and security.
"Users can work through a safe direct connector that keeps content in Box faster, more seamless, and fully governed."
Secure AI Integrations with Box MCP Server
Kamil showcased Box MCP Server, designed to enable AI systems to securely interact with Box’s enterprise-grade content. Kamil emphasized the server’s capabilities, including preserving strict security protocols and managing permissions to ensure users only access authorized content.
"Box MCP server fundamentally preserves Box’s enterprise-grade security and permissions, so users can only get answers from content they're authorized to access."
Box MCP Servers uses the streamable HTTP transport type, with an accessible endpoint at mcp.box.com. Authentication is handled via an already established OAuth 2.0 framework, ensuring a seamless experience for businesses and partners.
Partner Opportunities
Another focal point was the Box’s expanded partnerships with leading AI providers. Enterprises can now access the Box MCP Server in platforms such as Anthropic, Mistral AI, OpenAI, Azure, GitHub, and Amazon. Additional collaborations with Salesforce, Glean, IBM, and Workato, Groq are slated to go live soon, further embedding Box MCP into the AI ecosystem.
"You can now access the Box MCP server in platforms from partners including Anthropic, Mistral AI, OpenAI, Azure, GitHub, and Amazon, where Box is prominently listed in their MCP catalogs. Also, Salesforce, Glean, IBM, and Workato have announced support, and we are collaborating with them to bring it live in the upcoming months."
Live Demonstration: Real-World Applications of MCP
The live demo showed the integration capabilities of the Box MCP Server with AI tools like Anthropic’s Claude and OpenAI's Agent Builder. These examples illustrated how AI agents could effectively query Box tools for real-time enterprise data in a governed, secure environment.
For instance, Claude was shown retrieving content via Box MCP Server to generate a complete report based on enterprise data. Additionally, OpenAI’s Agent Builder demonstrated a workflow where files were extracted from a designated Box folder and used to generate a welcome text.
"Claude successfully checks for the available Box tools and utilizes them to get content from the enterprise... a report is generated... based on the data that it found from Box."
"OpenAI’s Agent Builder found the files in a workflow and generated a welcome text using the Box MCP server."
Use Cases Across Business Teams
The speakers illustrated practical applications of Box MCP across diverse business functions:
-
Sales Teams: Accelerating deal cycles by allowing AI agents to analyze product roadmaps stored in Box for customized proposal generation.
"Accelerating deal cycles by having an agent inside Salesforce that can automatically analyze the latest product roadmaps and Box, and generate tailored proposals."
-
Legal Teams: Streamlining contract review processes by enabling AI agents to instantly review entire documents and ensure compliance without manual upload.
"An agent like Claude can instantly review entire contracts stored within Box."
Developer Engagement and Open Source MCP Server
Recording clip here
Scott closed the session by highlighting the potential for developers to leverage the open-source MCP Server for expanded integrations. He also extended an invitation to connect with Box's Developer Relations team for further engagement and collaboration. Scott also compared and contrasted the open-source MCP server with the Box MCP server.
"Working quite closely with developers around adopting the Box MCP server, as well as the open-source community MCP server that we also have available."
Hurrey likened MCP to a "toolbox," empowering developers and AI agents with simple, scalable access to enterprise data.
"Think of MCP as the toolbox that AI agents—akin to mechanics—can use to access tools and data securely."
Questions for the team?
Please share in the replies!