Unlocking AI with Model Context Protocol (MCP)
0 likes79 views
Model Context Protocol (MCP) is quickly becoming the backbone for advanced AI ecosystems. The protocol enables context sharing, agent collaboration, and integration with large language models (LLMs). In this session, we’ll dive into what MCP is, why it’s gaining traction, and how it addresses the growing complexity of AI workflows. You’ll learn how MCP standardizes communication between AI agents and tools. Whether you’re building AI products, integrating models, or curious about the future of AI this session is for you. This session was given at Beer City Code 2025 in Grand Rapids MI on August 9th, 2025 by Brian McKeiver.
Unlocking AI with Model Context Protocol (MCP)
- 1. Unlocking AI with Model Context Protocol (MCP) by Brian McKeiver
- 2. 4x Microsoft Azure + 14x Kentico MVP Brian McKeiver, Co-Owner at BizStream, was awarded the title of Kentico MVP back in 2012, and has held the ever since. That’s 14 straight years. For his knowledge of building solutions in Microsoft Azure, commitment to educating the community, podcasts, and speaking engagements, he was awarded as an Azure MVP in 2022 and 2023. In 2024, Brian was also awarded as a Microsoft Azure AI MVP for his experience in the Microsoft Artificial Intelligence stack.
- 4. About BizStream We specialize in digital strategy, design, and custom web development, for enterprise clients across North America. We utilize our proven process, years of experience, and a variety of technologies to solve complex integrations and create custom software solutions. 2001 BizStream was Established 500+ Projects Completed 40+ Team Members 55+ Industry Certifications 4
- 5. Our Team is a Differentiator • Our team is based in West Michigan, near Grand Rapids. • We do not do any outsourcing, offshoring, or near-shoring, and haven’t for years. • Our Hybrid approach allows the team to gather at the BizStream Office and work from home based on flexible schedules. • We feel that this cohesion gives us a creative and collaborative advantage over fully remote agencies.
- 6. A sample of clients we work with
- 7. Making an Impact BizStream Academy is a non-profit organization that BizStream founded and maintains to this day. Our goal is to spark interest and awareness in youth about technology and general web development. We also contribute our time and resources to many other local area causes, such as the organizations listed to the right.
- 8. Enough About Me, What About You All?
- 9. Isn’t AI Smart Enough? Not really.
- 10. What is MCP? Model Context Protocol (MCP) connects AI models (like Claude, GPT-4o, etc.) to external tools and systems. That can be your app’s API, a product database, a codebase, or even a desktop environment. MCP lets you expose these capabilities in a structured way that models can understand. MCP isn’t a library or SDK. It’s a spec, like REST or GraphQL, but for AI systems and AI agents.
- 11. Made Up of Two Parts MCP Client A component that communicates with an AI model via the Model Context Protocol to send context, instructions, or tool requests MCP Server A system that implements the Model Context Protocol to manage, interpret, and respond to client inputs using AI model capabilities and contextual tools.
- 12. Why MCP? 12
- 13. Evolution of AI Ecosystems Problem: Typical AI connections between LLMs and real-time systems are getting complicated. • Generative AI / LLMs • Context • Retrieval Augmented Generation (RAG) • Function Calling / Tools • Memory While these AI patterns unlock powerful capabilities, they also introduce significant integration challenges. 13
- 15. The Challenge 15
- 18. Integration to the Rescue! APIs can solve this problem. Right?
- 23. Solution (MCP)
- 25. What is the difference between an MCP server and an AI Agent? An AI Agent is the model (client/system using the model) that plans and takes actions. An MCP server is the environment that defines what actions are possible. The agent decides what to do. The MCP server tells it how to do it. Think of it like this: • The chef decides what to make (AI agent) • The kitchen determines what tools and ingredients are available (MCP server) • The chef can only work with what the kitchen exposes You need both. The agent figures out the intent. The MCP server makes sure it stays on track and uses approved tools.
- 26. 26 Some Demos
- 27. MCP Demo #1 ss
- 29. MCP Demo #2
- 30. Not Everything is Perfect with MCP
- 31. Pace of Innovation and Change is Scary MCP Security Risks • Prompt Injection • Tool Poisoning • Privilege Abuse • Tool Shadowing / Shadow MCP • Authentication Bypass • Rug Pull Attacks • Command/SQL Injection & Malicious Code Execution • Sensitive Data Exposure & Token Theft Mitigation Tips • Sanitizing inputs to prevent prompt injection, enforcing strict auth protocols like OAuth, applying the principle of least privilege across agents and tools. • Configuration audits, RBAC enforcement, and strong tenant isolation. • Secure credential handling, including encryption and token rotation.
- 32. References
- 35. 35 MCP Turns Your AI from a Chatbot to a Workflow Automation Powerhouse
- 36. BizStream’s AI Practice Free initial consultations on building intelligent solutions with AI for your business and applications. Come check out or booth! www.bizstream.com/contact Concept Discovery, Research, Design, Proof of Concept offerings. Build Building intelligent AI apps and solutions in .NET / C# / JavaScript / Python. Evolve Continuous improvement, ongoing support, and optimization.
- 37. Whatever It Takes Work Hard, Play Hard We are a Team Be Fearless Foster Growth in Others Care Be Positive Thank you. Go Whitecaps!