By Vincent Hsu, IBM Fellow, VP & CTO, IBM Storage
Until recently, it’s been difficult for AI systems to communicate or collaborate with the different tools, data sources, and services they use, which has limited their value in the enterprise.
That’s why, less than a year after its introduction by Anthropic, the Model Context Protocol (MCP) has already become a key part of modern AI infrastructure – it enables AI systems to maintain and share rich, persistent contextacross sessions, tools, and even different models.
IBM believes that it’s time for AI tools to evolve from brilliant but isolated consultants into cohesive and reliable teams, and that MCP is the bridge needed to move from today’s fragmented experiences toward more seamless, continuous, and intelligent AI interactions.
Today, I’m excited to share a foundational step on that journey, with an initial rollout of the open-source IBM Storage Insights MCP Server codebase on github.
Why All the Buzz About MCP?
AI tools are already having major impacts in all industries. But the evolution from chatbots to agents is highlighting a major challenge – we ask these systems to analyze unprecedented amounts of information and make recommendations, but we often don’t trust their output.
The result is that AI tools have been largely confined to an advisory role, rather than being used to autonomously orchestrate diverse resources to best achieve defined outcomes.
Model Context Protocol (MCP) is gaining attention because it enables AI systems to maintain and share rich, persistent context across sessions, tools, and models. This means users don’t have to repeat themselves, and AI agents can collaborate more effectively with continuity and personalization. By standardizing how context – like user goals, history, and preferences – is stored and transferred, MCP unlocks smarter, more seamless interactions and lays the foundation for multi-agent ecosystems and truly useful personal AI.
Leading AI companies are rapidly embracing MCP to standardize how models connect with tools and share information. OpenAI added MCP support to its ChatGPT desktop app, Agents SDK, and API in early 2025, enabling easier integration across apps. Google DeepMind followed, calling MCP a key open standard for the future of AI assistants, and Microsoft integrated MCP into Copilot Studio to streamline tool access and workflow automation. Dozens of other platforms – from Stripe and Replit to Cursor and Neo4j – are adopting MCP to enable smoother collaboration between AI agents and real-world systems.
Getting Started – Enhancing Data Observability in IBM Storage Insights
In looking for a suitable use case for prototyping these new capabilities, our development team identified the significant customer benefits that would result from making our flagship observability platform, IBM Storage Insights, MCP-aware. The result is the open-source MCP server we’re making available today, which connects directly to Storage Insights to allow AI assistants to securely access the rich, real-time data streams they need for AI reasoning pipelines.
The power of any AI agent is directly proportional to the quality of the context it receives. The Storage Insights MCP Server is uniquely powerful because it draws from one of the industry's most mature and comprehensive observability platforms. It doesn't just expose basic array health; it allows customers to leverage years of their organization’s historical performance data, workload-specific metrics, capacity trends across hybrid cloud environments, and deep insights into system components.
While the broader vision of a fully autonomous data center is a multi-year journey, the benefits of this agent-based approach are tangible today. For example, MCP can make the daily life of a storage administrator less stressful and more efficient.
Your AI “Morning Coffee” Report
For an enterprise storage admin, the MCP server we’ve posted today provides the potential for a different kind of work day, one that begins with a simple, conversational request: “Good morning. Can you give me my “morning coffee” report?”
In response, an AI agent, leveraging the MCP Server to query your environment via Storage Insights, can deliver a comprehensive, prioritized summary. It can instantly tell you that you have ten storage systems under management, with three currently in an error state and one in a warning condition. It can highlight the most critical issues, such as probe failures on two of your XIV systems or that your IBM DS8OOO system is nearing its capacity limit at 74% utilization.
[Caption: At left, the reasoning displayed as an AI assistant runs the “Morning Cup of Coffee” query, with an output dashboard at right.]
This isn't just data retrieval; it's synthesis. The agent intelligently surfaces the critical alerts that require your attention, separating them from the noise.
More importantly, its output isn’t static – you can follow up by asking the agent to visualize complex data: “For that DS8000 system, draw a line graph of capacity metrics for the last 20 days” The agent can then generate an interactive chart showing the steady growth trend, highlighting periods of especially rapid increase over that period, and recommend that you consider expansion before reaching the 85-90% threshold.
This moves the administrator from data collector to strategic decision-maker, dramatically reducing the mean time to resolution for both performance troubleshooting and capacity planning. It helps AI agents move beyond simple “what is” questions to complex “what if” and “why” analysis, forming the basis for true, data-driven autonomous operations.
Next Steps – From Queries to Autonomous Operations
This initial phase is focused on providing read-only observability, but we see many areas for future development of the open source code base, with such possibilities as enterprise internalization, hardening the server for multi-tenancy and policy control, transactional capabilities, and ultimately a full-fledged conversational AI assistant layer.
We believe the path to more intelligent and autonomous operations must be a collaborative one, built on open standards and broad community participation. We are in the early, exciting phases of this transformation, and we are not doing this alone. By contributing to this open standard, we invite you to participate not just in a project, but in a mission to define the future of intelligent infrastructure.
For those ready to dive into the technical details, a more hands-on, step-by-step guide is available in a related blog post and within the documentation of the IBM Storage Insights Open Source MCP Server codebase itself.
We invite you to download the server, review the documentation, and begin experimenting today. The full source code is available now in our new GitHub repository:
Engage with the Community- Join the IBM TechXchange- Global Storage Forum