with Ed Anuff and Anant Jhingran
Are MCPs the new APIs?
April 22nd, 2025 | 44:03 Runtime
Episode Transcript
TIMESTAMPS
00:01:24 - Exploring Large Language Models and Tools
00:05:07 - The Rapid Evolution of Technology
00:06:57 - Understanding MCP and Its Importance
00:12:05 - The Role of APIs in Tool Integration
00:17:48 - Challenges in API and MCP Implementations
00:19:34 - Crafting Effective Tool Descriptions
00:25:17 - Architectural Considerations for MCP Usage
00:27:51 - Navigating API Management Challenges
00:36:05 - Testing MCP Servers with LLMs
00:44:43 - The Continuum of Procedural and Agentic Approaches
00:51:27 - Future Directions in AI, Data, and APIs
QUOTES
“You've gone from the realm of, oh, my God, look at what it can do, to, can I actually bet my business on something? And that gap between, wow, and I can bet my business, is a gap that I think we need to address.”
“And how do you know that the information that's passed, that is gotten from one MCP server is passed reliably into the second one? So we have been just kind of thinking deeply about whether you want your APIs or MCP servers to do really complex things or really simple things.”
“So how do we think about testing our implementation of MCP servers, which is no longer just about does one MCP server do the right thing? But does the LLM chain together its own CoT, chain of thought reasoning and connect with MCP servers?”
“There are a lot of parallels there that we saw, which were that you had a new type of applications being built that were leveraging APIs. And now we are at a point where people are building these agentic applications and they're figuring out how to connect them to their data and apps, and particularly within enterprises that we spend a lot of time working with and trying to sort of separate out, what is the hype from what is fact.”
“Both you and I are big proponents that you should learn by doing. And the advice that you and I have given many folks in IT and, and in business and enterprises over the years is that you need to go and get your hands dirty in building these things.”
“Good API design makes for great tools. Bad API design makes for okay tools.”
TIMESTAMPS
00:01:24 - Exploring Large Language Models and Tools
00:05:07 - The Rapid Evolution of Technology
00:06:57 - Understanding MCP and Its Importance
00:12:05 - The Role of APIs in Tool Integration
00:17:48 - Challenges in API and MCP Implementations
00:19:34 - Crafting Effective Tool Descriptions
00:25:17 - Architectural Considerations for MCP Usage
00:27:51 - Navigating API Management Challenges
00:36:05 - Testing MCP Servers with LLMs
00:44:43 - The Continuum of Procedural and Agentic Approaches
00:51:27 - Future Directions in AI, Data, and APIs
QUOTES
“You've gone from the realm of, oh, my God, look at what it can do, to, can I actually bet my business on something? And that gap between, wow, and I can bet my business, is a gap that I think we need to address.”
“And how do you know that the information that's passed, that is gotten from one MCP server is passed reliably into the second one? So we have been just kind of thinking deeply about whether you want your APIs or MCP servers to do really complex things or really simple things.”
“So how do we think about testing our implementation of MCP servers, which is no longer just about does one MCP server do the right thing? But does the LLM chain together its own CoT, chain of thought reasoning and connect with MCP servers?”
“There are a lot of parallels there that we saw, which were that you had a new type of applications being built that were leveraging APIs. And now we are at a point where people are building these agentic applications and they're figuring out how to connect them to their data and apps, and particularly within enterprises that we spend a lot of time working with and trying to sort of separate out, what is the hype from what is fact.”
“Both you and I are big proponents that you should learn by doing. And the advice that you and I have given many folks in IT and, and in business and enterprises over the years is that you need to go and get your hands dirty in building these things.”
“Good API design makes for great tools. Bad API design makes for okay tools.”