Real World Serverless with theburningmonk #116: AI Agents, MCP and the problems with AI benchmarks | ft. Matt Carey
7 snips
Apr 19, 2025 
In a fascinating discussion, Matt Carey, a founding AI engineer at StackOne and AI Demo Days founder, breaks down the complexities of AI agents and the MCP protocol. He explains the challenges of AI benchmarks, emphasizing the disconnect between user experiences and their results. Matt also tackles the importance of curated content in navigating the AI landscape and shares insights on networking within the AI community. His expertise sheds light on effective strategies for leveraging AI technologies, making the conversation a must-listen for tech enthusiasts.
AI Snips
Chapters
Books
Transcript
Episode notes
Matt's StackOne Integration Journey
- Matt Carey joined StackOne to help build API integrations faster and create a workflow builder for internal and customer use.
- They achieved 90 integrations in three months, significantly faster than competitors, and are now focusing on AI agents using their unified integration system.
MCP as a Universal Plugin System
- MCP is a universal plugin system for AI clients, like Chrome extensions for browsers.
- It lets users extend AI application functionality by adding tools they want, empowering user control and customization.
MCP API Capabilities
- MCP defines three APIs: tools, resources, and prompts, but currently only tools are widely supported.
- Tools allow AI clients to run code that interacts with external systems, enabling diverse creative and functional extensions.
