"One of the things that I'm really excited about right now is that a lot more people are starting to recognize that MCP isn't just this protocol for an agent or some other tool to talk to a set of APIs, but can be a protocol for agents talking to other agents."
The Model Context Protocol (MCP) is evolving beyond local developer experiments and into the secure, remote infrastructure that will power the next generation of the internet. Brendan Irvine-Broque, Director of Product at Cloudflare, joins us to share a roadmap for this future. He explains how Cloudflare's "customer zero" philosophy of dogfooding their own tools provides a unique perspective on what it takes to scale MCP for production.
Brendan makes the case for observability as the ideal starting point for enterprises and lays out the vision for MCP's ultimate destination: a universal protocol for agent-to-agent communication. The conversation explores how remote servers can create a decentralized layer for security and user memory, and what the exciting development of MCP UI means for the future of chat-based applications. This is an essential look at the next wave of agentic systems and the infrastructure required to build it.
Show Notes
- Learn more about Cloudflare's work with AI: agents.cloudflare.com
- Read the latest from Cloudflare: The Cloudflare Blog
- Connect with Brendan Irvine-Broque: X @irvinebroque | LinkedIn
- Cloudflare's Unique Primitives Mentioned: Durable Objects
- The MCP UI Project: MCP UI on GitHub (Project by Ido Salomon)
- Observability Tools Mentioned: Datadog | Honeycomb
- AI Tools Mentioned: Block/Square's "Goose" | Cursor
Transcript
(Disclaimer: may contain unintentionally confusing, inaccurate and/or amusing transcription errors)
[00:00:00] Andrew Zigler: Today we're talking to Brendan Irvine-Broque the director of product at CloudFlare, a company that needs no introduction, by the way, because it runs the internet at scale and has a front row seat to the way things like MCP are hitting big engineering orgs. And back in March on Dev Interrupted, we covered Brendan's article actually in a news segment, an article about building and deploying remote MCP servers, the CloudFlare, and it opened a whole new, world for us.
[00:00:24] Andrew Zigler: As you know, we've been talking about MCP on Dev Interrupted for a little bit now, and I've been tinkering with it as well. MCP server adoption is dominating team agendas. And why Brendan's hearing it from everyone that he talks to in leadership right now, and what does it even mean to get something like that,
[00:00:41] Andrew Zigler: right? So, Brendan, welcome to Dev Interrupted.
[00:00:45] Brendan Irvine-Broque: That's a wonderful introduction. Thanks first so much for having me on here.
[00:00:48] Andrew Zigler: Great. Well, we're really excited to have you here. And, and I wanna kind of start by talking about CloudFlare itself because it's in a really unique position and it operates as like really the ultimate proving [00:01:00] ground for anything that's gonna make it on the web. at CloudFlare, I imagine y'all are dogfooding internet at scale and any tool that succeeds and works for your teams internally has the potential to transform the entire internet. And that's a really high stakes for internal adoption and a story everyone would be curious about. So, uh, I'm wondering when you introduce a powerful new technology like MCP inside of CloudFlare. What's like a guiding principle at CloudFlare that you come back to?
[00:01:31] Brendan Irvine-Broque: Yeah, so what we talk about at CloudFlare so often is when we build products, we're building for like our own internal customer, zero. First. We're building things that are part of our own platform, but we use our own platform. Form ourselves, and that's what, you know, that internal dog footing is what makes it good for our own customers.
[00:01:49] Brendan Irvine-Broque: I wouldn't feel good as a, you know, as a product leader, if I'm telling a customer of Cloudflare's that they should go and adopt something if, you know, we're not really using it ourselves. And so we always come [00:02:00] back to that and it's, for example, internally we use a lot of our own MCP servers when teams are debugging their own CloudFlare workers and they want to ask observability questions about what's going on.
[00:02:12] Brendan Irvine-Broque: And that kind of feeds back into our own teams and our own product development and what we need to kind of make better for our external customers as well. There's kind of just something to like, engineers on a team, being able to talk to really easily to engineers who are, who are building that platform feature and kind of like go back and forth really, really quickly in fast cycles.
[00:02:30] Andrew Zigler: I definitely wanna talk a little bit more about observability in, in this world with you, uh, especially as we kind of start to guide along the MCP story here. but I'm, curious like. CloudFlare, where would you say the spark began for MCP in using it? Like was there maybe like an experiment that somebody spun up or was it a strategic top-down directive?
[00:02:51] Andrew Zigler: What did it look like?
[00:02:53] Brendan Irvine-Broque: So for us, you know, we're obviously paying attention kind of ears to the ground on. What did developers wanna build? [00:03:00] And I remember it was almost a year ago now, kind of hearing about MCP coming out of Anthropic and this kind of new, new protocol for kind of agents talking to, to APIs and to other agents.
[00:03:10] Brendan Irvine-Broque: And so, you know, we got excited because a lot of what we see our customers doing with CloudFlare workers is, you know, they're building some bit of software that needs to talk to some other API and service and people are starting to build. AI agents on us and exploring those different things with durable objects and, and all kinds of different things that are possible on, on CloudFlare.
[00:03:31] Brendan Irvine-Broque: And it was funny because initially, you know, we had some ideas ourselves about like, how should different services be able to communicate with each other and how might that work? And when we saw MCP come out, we said like, okay, well there's a lot of interesting people behind this and maybe there's a thread to pull on.
[00:03:47] Brendan Irvine-Broque: And it was all local at the time and we kind of hacked some different things together and, you know, there are some people who are interested in MCP, but it, but it didn't really blow up until probably the first couple months of 2025. And really for [00:04:00] us it was just about meeting the moment of like what developers wanna, wanna build.
[00:04:04] Brendan Irvine-Broque: First and foremost, it was maybe less at that point about like internal adoption and different things of just like there's so much excitement from, from people in the community.
[00:04:14] Andrew Zigler: Yeah, I, I can really relate to that because what I saw with MCP when it hit the scene it it became like a new primitive on which everyone could build in the same way that AI was a new primitive where everything and anything before an engineering could. with AI and it would radically change what that was before.
[00:04:30] Andrew Zigler: Everything that went through it was different. MCP was the same 'cause it's modular. And there became an MCP for everything, right? You had this sudden proliferation once, you know, like you said earlier in 2025 when it started to catch on, you started to get these MCP server directories and a repo for every kind of MCP connector that there was.
[00:04:48] Andrew Zigler: And you know, everyone has read. The articles, uh, on medium of, of folks being like, you can't just map one-to-one your API to MCP, but that doesn't mean, no one did. So all of the APIs, that [00:05:00] do do exist, they, they, they've kind of been started to push through this MCP world, right? I think you're right though that it comes down to like finding like the value and, and the usage and, and you saw it as like, oh, wow, there's something to experiment with here.
[00:05:13] Andrew Zigler: There's like a, a building block that people can use. And that's what CloudFlare, I think is looking for. Like with things like durable objects and the whole CloudFlare ecosystem is about the building blocks to deliver the web. So MCP was evolving as one of those.
[00:05:27] Brendan Irvine-Broque: I mean, at the end of the day, what people are trying to do, right, is that they're trying to figure out as businesses. How do they fit into all of these different kind of AI clients, whether it's ChatGPT or Claude or Cursor or whatever's gonna come next and re recognizing that, that that what you build when that is the interface that your customers use looks a lot different than a website with its own interface.
[00:05:51] Brendan Irvine-Broque: You know, the, uh, in some sense, I mean, this is now something that I feel like everybody says, but at the time we were saying. Saying it ourselves was like, these different AI [00:06:00] tools are like the new web browsers. You know, you go first to chat GPT, you don't go to Chrome. And I think at the time the only thing we believed in strongly was like, I don't know, the future probably isn't just more and more and more Chrome tabs.
[00:06:11] Brendan Irvine-Broque: I think we all hit a breaking point in our browser around that. And just trying to help internet businesses and companies figure out how they, how they plug in and. The off piece of MCP, especially initially was really the hardest hurdle for people to get through if we wanted to make that leap early on.
[00:06:28] Brendan Irvine-Broque: From, this is all local and really only for all of us software engineers and developers to run ourselves and, and make something that was truly accessible that, you know, everyone in a large organization or consumers can use.
[00:06:42] Andrew Zigler: Right. And accessible and secure. Something that was built more to scale. And it, it sounds like that was the, like a signal, right? Like a milestone to MCP that you had to, cloudFlare had to meet that mission first of making it secure, making it something that could work remotely and in a server. Right? And that [00:07:00] multiple folks, man, uh, connected to a managed service has a whole different, uh, world to it than the MCP Lo, uh, servers that live locally.
[00:07:07] Andrew Zigler: Right? So, was, it was that, I guess the, the movement within CloudFlare, you, you are able to do that and then you can say, okay, maybe this has legs. Now how do we scale this to the rest of the internet?
[00:07:21] Brendan Irvine-Broque: Yeah, I mean, one of the things that we picked up on initially was. Especially in the first iteration when of the MCP spec, it was actually pretty hard to figure out how to manage this kind of long lived connection that an a client has to have with an MCP server. And one of the unique things about Cloudflare's developer platform is that, you know, we have this really interesting primitive, we call durable objects, which you can kind of think of as these like.
[00:07:47] Brendan Irvine-Broque: Stateful workers that, you know, they're these singletons and so you can, you know, create an infinite number of these and you can connect to them. And it, we saw this like easy kind of way to say, okay, well maybe [00:08:00] elsewhere it would be hard for each session with an MCP server to kind of have its own instance.
[00:08:05] Brendan Irvine-Broque: Its that it's connecting to, but there was just this like very natural fit for us. One of the things that I'm really excited about right now is that a lot more people are starting to recognize that MCP isn't just this protocol for like, you know, an agent or, uh, some other tool to talk to a set of APIs, but can be a protocol for agents talking to other agents.
[00:08:29] Brendan Irvine-Broque: And the reason that that was kind of this thing that hit us over the, the head was, well, our model has always been with the things that we build with agents, that they're all, you know, you can create an infinite number of instances of agents and, and it's just a kind of like, naturally fit, fit, like a glove.
[00:08:45] Brendan Irvine-Broque: And so I'm, I'm really excited about that future of like MCP as a protocol for agents to communicate.
[00:08:52] Andrew Zigler: I completely agree with you. I'm, I'm familiar with durable objects. I've used them before and that's the exact mapping that I see too, is like, wow. Uh, talk about [00:09:00] how Cloudflare's already built to deliver that exactly how it was. And so it was just really cool to see that internal journey and how you all meet the mission and, and when you take that focus outward, you know, you're, you're, you yourself are in this really unique position where you take all of this product knowledge and this really global domain.
[00:09:18] Andrew Zigler: Knowledge from CloudFlare, and you bring these to these engineering leaders who, who meet with you, and when, when you do sit down with them, you, you see a lot of patterns emerging. That's what you and I talked about in our initial conversation patterns between all of these leaders, the questions they ask, the concerns that they have. So what do you think is like the number one question burning in all of their heads right now?
[00:09:38] Brendan Irvine-Broque: so many of the engineering leaders I talked to have done a wonderful job of kind of spinning up these tiger teams or small groups of engineers who are. Whether they're creating agents or MCP servers or different ways that their company can use ai, and what they're trying to do is they're trying to say, well, how am I gonna give this to a really big org where I, you know, don't know [00:10:00] everybody, and maybe there's a new hire who's gonna show up next Tuesday.
[00:10:03] Brendan Irvine-Broque: And how do I do that safely and how do I also just understand what people are using or getting value out of, or kind of roll this out securely? we've heard that over and over and over. That's just like such an interesting, consistent theme is like, you know, I think engineering leaders really get it, that this is what's gonna accelerate their organization.
[00:10:22] Brendan Irvine-Broque: And so we spent a bunch of time kind of building kind of MCP portals back into Cloudflare's set of zero trust products to be able to give people these tools to kind of administer a set of MCP servers to their team and kind of create this gateway that they can hand over and say. Here's the, here's a set of tools that, you know, maybe me as an engineering director, I want this set of teams to be able to use and maybe there's a different set of tools that, you know, a different group of users, I, I wanna make sure that they have access to.
[00:10:52] Brendan Irvine-Broque: So that kind of rollout and management of everything is a big problem in this space. And so I'm, I'm excited about the things that we're trying to, trying to do for people that.
[00:10:59] Andrew Zigler: [00:11:00] Yeah, the, the rollout is a challenge because part of it is, is the technology. Another big part of it is like the people and the communication process behind it. And I like that you called out that everyone's done a really good job at Tiger Teams. I, I think that is a universal cross of where everyone has identified the opportunity.
[00:11:16] Andrew Zigler: I. Fostered the, that curiosity hopefully at this point, and have assembled those first movers, those, those heavy users, those early adopters, and gotten them to kind of start pulling the cart forward. Right? And, you know, our, our own company, we have those internally. Many of our listeners, they, they relate with the story too. But of that, like, you know, you get that cluster of people who are very much a aboard and can see it. you know, getting more people on board is a. Can be more difficult, especially when you're dealing with folks with all different levels of AI interaction and comfortableness, you know? So what, do you see teams get stuck anywhere?
[00:11:50] Andrew Zigler: Or are there things that.
[00:11:51] Andrew Zigler: jump out at you as opportunities?
[00:11:53] Brendan Irvine-Broque: I mean, one of the most. Basic ways that people get stuck is because in some sense [00:12:00] progress has happened so much faster than I think many people would've expected. With so many MCP servers getting built, and a lot of these servers expose so many different tools, it's really easy to say, well, you know, I need access to this, this, this, this, and this, and install a bunch of things, and suddenly you've.
[00:12:16] Brendan Irvine-Broque: Blown up the context window. Uh, LLM is quite confused 'cause you've given it 25 tools that all start with the word update. And you know, somebody comes along and maybe they're less familiar with context engineering and you know how this stuff works behind the scenes and they're just like, it doesn't really work for me.
[00:12:34] Brendan Irvine-Broque: I'm not getting the results that I want. And so, you know, a lot of what I see people going through right now who are building MCP servers is starting to rethink like, what tools should or shouldn't be available? Maybe we've gone overkill, maybe we've exposed too many things or, uh, more nuance in terms of like which tools I actually do wanna enable for certain use cases in different contexts or groups of people.
[00:12:58] Brendan Irvine-Broque: You know, I think that everybody, [00:13:00] everybody, every team wants to move really fast with ai and a lot of people made these big pushes to get MCP servers out the door. And now we're just in this, this cycle of people saying like, well. How do you measure the success of A MCP tool? And when you think about that, it's actually a really, really hard problem of like, oh yeah, the tool runs.
[00:13:18] Brendan Irvine-Broque: Does it give you a response like, yes, it's available, but like, is that response good? Well, I can pull on this thread and say, well, I can do my own evals for it, but I'm dreaming up how I use it. You might use it in a totally different way. Like evals for tools are remarkably hard, especially when you don't control the client.
[00:13:37] Andrew Zigler: Yeah, you've called out a really good thing here. We talked about a bit on, on, on Dev Interrupted as well about evals in the world of, uh, of understanding how AI is used and how, you know, they're, they're a helpful tool, but they're just one part of, of the picture. And there's other things too, like, more of an observability level of understanding even what was the data that went into it, making that ultimate response that it did.
[00:13:58] Andrew Zigler: Because in many [00:14:00] cases you have MCP servers that are maybe connecting to data platforms and surfacing information, and they're doing that with tool calls. Sure. But then like, you know, what data did it actually get where that data come from in places with a lot of data? That is a huge question. So, there's all these different parts that have to be like very. carefully utilized. And also like that you called out that if you have a bazillion tools, like it's not gonna work. I have experienced this. If, if you are listening to this and you are using MCP servers and you have a whole bunch of them hooked up to something right now, like cursor or Yes.
[00:14:30] Andrew Zigler: Code. I'm begging you, please go in there and turn some of them off or, or understanding, understand how many tools that you are exposing to your LLM at any given time. 'cause it really does. Up the level on confusion. and, and so like all of these little pitfalls, right? We're learning them all together as an industry in real time.
[00:14:47] Andrew Zigler: We're all writing the blog posts about them. We're all rediscovering those engineering principles underneath. And, many leaders are building stuff on top of this too, and. It sounds like maybe there's a, a framework that folks [00:15:00] should bring to it when making decisions around MCP adoption and it, maybe those are things that are rooted in like security or compliance, but I'm just wondering if there's any like big themes that jump out at you.
[00:15:11] Brendan Irvine-Broque: One that I've seen once kind of category of MCP services, I've seen work remarkably well is. Been everything around observability
[00:15:20] Andrew Zigler: Hmm.
[00:15:21] Brendan Irvine-Broque: You know, one of the classic kind of concerns that engineering leaders I've had when I've kinda run engineering teams before is you maybe have four or five different tools and systems for how you get logs, traces, metrics, data.
[00:15:35] Brendan Irvine-Broque: You know, you use like a Datadog, a Honeycomb, a CloudFlare, all these things.
[00:15:39] Andrew Zigler: Right.
[00:15:40] Brendan Irvine-Broque: And how do you pull this stuff all together? You know, you go to debug things in an incident and you're saying like, well, I gotta look, pull up six or seven different tabs in Chrome and like, correlate different things just to figure out like, okay, it's this.
[00:15:54] Brendan Irvine-Broque: And, you know, I've talked to so many people who have found success just saying, [00:16:00] here's an MCP server where I can query kind of logs or metrics or data from one system and I can bring in the other MCP server and I can have, an agent just look at correlation and look at what might be happening here and piece the two things together.
[00:16:16] Brendan Irvine-Broque: And the reason I bring it up is. A that it's just a successful use case, but it's also pretty universal to engineering teams. Everybody has to debug things and it's reasonably focused in that like, you know, you're not trying to roll out all of the possible things that people could do with tools and AI at once.
[00:16:35] Brendan Irvine-Broque: And it's fundamentally, you know, it's read only. And so some of the things that we're worried about of, you know, exposing a tool that can then mutate and change data or do some dangerous operation, you know, somewhat start to fade and, you know, if you've done the right sanitization of logs for PI and other things, like you're, you're actually in a pretty good, good spot.
[00:16:55] Brendan Irvine-Broque: And so that's the, that's the use case that I'm just seeing, like work pretty consistently outta the box.[00:17:00]
[00:17:00] Andrew Zigler: Yeah, and, and it's taken us time, obviously as a. an industry to even get there. Right? Because MC P's been on the scene now for about a year or so, like we're coming up on it, being in, in, in the scene for a little while, but it's gone through a lot of phases of evolution in that time and adoption and, and becoming more mainstream. And you know, something that you mentioned in our initial call is a lot of folks that were early to the MCP scene and building stuff on top of it and making tools with it. Then would have to maybe like reset. Or, or, or kind of go back to the drawing board once they tried to take it out to more broadly their teams or however they wanted to ship it.
[00:17:33] Andrew Zigler: So what was like the most common, like, you know, we got a wrong moment that you saw from your perspective in that world.
[00:17:40] Brendan Irvine-Broque: I mean, you know, the classic thing of, translate a API schema one-to-one into a bunch of tools. I mean, that's the, the clearest one. think, you know, there's, there's interesting nuances to this, right? Where it's like, uh. forgetting about the importance of descriptions. You know, at the end of the day, uh, we're [00:18:00] also used to, 'cause we're, we're engineers, we're programmers being like, okay, well the method is called this.
[00:18:05] Brendan Irvine-Broque: And so like, yeah, yeah, I need to leave like a js dot comment that describes stuff. But like half the time I don't really remember to do it that well. Like to the model that's like the information that tells you what you're supposed to use this for and.
[00:18:20] Andrew Zigler: Yeah.
[00:18:20] Brendan Irvine-Broque: you the context. It's literally like, it's in the context window.
[00:18:24] Brendan Irvine-Broque: And you know, some of the, some of the patterns that I've seen people try and it's worked pretty well, is like I've seen people develop some interesting mini evals frameworks to say like, how would I. Iterate on my tool descriptions to figure out what's actually gonna get me the best results. And just really thinking about the fact that I'm not, I'm not taking a set of APIs and exposing them just as like APIs, the way that my programmer brain would work.
[00:18:50] Brendan Irvine-Broque: But like, I have to think about, there's a model at the end of the day on the other side of this, and what is it, what is it doing with all of this?
[00:18:56] Andrew Zigler: Yeah. It's almost like there was a, there was a call to action where people [00:19:00] learned in there that it was about combining those API calls together into. What made the most sense? It was about, you know, in, in the old world of traditional engineering, where you would maybe have that function, that update something in a person's cart, maybe it fetches the cart in the first place and even fetches info about the user.
[00:19:19] Andrew Zigler: Maybe it updates, maybe it has to fetch the product list and then filter out probably like maybe there's a whole bunch of API calls in there. If you map that one to one, we all learned that. Now you're asking the LLM to successfully thread four or five tool call interactions together. Maybe like you said, you have a. 10 tools enabled, and they all have an update function in there somewhere. So it's trying to pick even the right tool, uh, in the toolbox to pick up. So, kind of dialing up the determinism. Is really important, and that's what MCP gives us with ai. I think that's why it's a powerful, powerful primitive. And then additionally, and this is kind of where I I I'm curious to lean on your perspective, your expertise here at CloudFlare is, you know, you bring it [00:20:00] into, you bring it into the infrastructure.
[00:20:02] Andrew Zigler: You make it remote, you put it behind authentication, you put it somewhere where it's a managed service and now suddenly you can see everything. It talks to everything that goes in and out of it. All the tool calls, all of everything that it learns and understands. And what does that level of observability like give you that people who are maybe only up until now have experienced local MCP servers, they don't even know about it.
[00:20:23] Brendan Irvine-Broque: Mm, that's interesting. I guess, you know, there's, there's two layers to this. One is. What is the inference that's happening? You know, in, uh, at CloudFlare we have a product called AI Gateway that lets kind of sits in between an application and kind of upstream inference, whether that's running on CloudFlare or that's running on, you know, open ai, open ai, anthropic, et cetera.
[00:20:47] Brendan Irvine-Broque: So it gives you like one layer. And there's other platforms that do this as well. Of like, what's, what are the inference calls that are happening? And then there's like, well, what are the tool calls that are happening? And you kind of need both to paint the [00:21:00] full picture of what happened and to understand like, you know, kind of like almost session replay what's going on.
[00:21:07] Brendan Irvine-Broque: And uh, I think that once you start to look at that and see real world use cases. The thing that I think we're seeing most now is how many times we're going back through the age agentic loop. How many times we're going and saying like, we're gonna run inference. We're gonna call a tool, we're gonna come back to the model.
[00:21:28] Brendan Irvine-Broque: The model's gonna think for a second, go call a tool, and we're just gonna go over and over. And that's great. 'cause you know, these models are phenomenal and amazing, but they're also, you know, takes time. It's the meaningful cost. And there's a lot of really, really interesting people, I think right now, this summer, who are kind of pushing at the edges of what might be different compute models here.
[00:21:53] Brendan Irvine-Broque: Like, there are agents that write their own code. There's all kinds of approaches to this that may. Make this just [00:22:00] way, way, way faster or way, way, way less expensive to run.
[00:22:05] Andrew Zigler: So, know, remote MCP being something I I think people are still getting used to and, and, and learning to adopt. Right. And there's a lot of like, questions that it solves around taking something like this to production or even letting folks like your customers access it. Maybe in a read only way, to use your data as a primitive for their own AI workflows and tools.
[00:22:24] Andrew Zigler: All things that we're experimenting with and that matter right now to, to engineering teams that are shipping products used by engineers especially. So like. Are there maybe less obvious, there's obvious security advantages to making it a managed service, but are there less obvious ones that like at CloudFare, that y'all have become more aware of as people pull it out of the local computer grasp and put it somewhere remotely?
[00:22:48] Brendan Irvine-Broque: I mean, one that I think is a little bit underappreciated is. Almost everybody who's built one of these things to date, what they're doing at the end of the day is they're taking an existing [00:23:00] API that they walked in the door with that already existed coming to 2025, and they're putting in some MCP server, some layer in front that sits between, you know, a client and that API.
[00:23:11] Brendan Irvine-Broque: And you know, that's why this stuff has grown really fast. Says that you can do that, you can just kind of retrofit this existing piece of software. And 2025 is the year of, OAuth, RFCs becoming the hottest topic in the world. Um, I have read more this year than any other year in my life combined. And, uh.
[00:23:32] Brendan Irvine-Broque: What gets what gets fun is you realize like, wow, there's all of these amazing specs that, you know, people have been fighting for, for, for years and trying to make happen, but maybe there wasn't the, the use case to push it over the edge around like, you know, finer grain authorization, et cetera. And most of these downstream APIs that people already have don't really support as fine grain authorization is.
[00:23:54] Brendan Irvine-Broque: You know, we pro company building them probably would want in this age of AI where this stuff really starts to [00:24:00] matter.
[00:24:00] Andrew Zigler: Yeah.
[00:24:01] Brendan Irvine-Broque: I say all that because this middle layer of an MCP server, it actually sits in an interesting role where. If you wanted to enforce a different kind of more granular form of authorization, you can do things like that.
[00:24:16] Brendan Irvine-Broque: Or you know you're building on top of an existing API you offer and it doesn't offer the right granularity of scopes. You could enforce some of that layer within an MCP server. Or enforce different types of guardrails and if there's a dangerous operation, and you can kind of hedge against that.
[00:24:34] Brendan Irvine-Broque: So there's, there's some interesting threads to pull on there where you can kind of sit in the middle. The other one that, we've done some cool demos around, uh, Dina Lov, who's on my team, I work really closely with, she did a great presentation at the, uh, MMCP conference, uh, a couple months ago about MCP servers that have their own memory.
[00:24:54] Brendan Irvine-Broque: And can remember what a user has asked them in the past. And this is one of these questions [00:25:00] that. I think is, is getting asked by everybody who's investing in AI companies, right? Like our is is the user's memory just gonna be a thing that open AI controls or anthropic controls? And there's some great ideas both like within our own developer platform with, uh, startups that are built on top of CloudFlare, uh, that kind of let people bring memory.
[00:25:21] Brendan Irvine-Broque: Across platforms and persist that over time.
[00:25:24] Andrew Zigler: Uh, that's really interesting. It, it really highlights how as a layer remote MCP servers become almost like a decentralization layer that allows people to control that bespoke experience that they bring with them. They plug into other tools that maybe remembers things from them in other sessions, or how they use things or what they're authenticated for.
[00:25:48] Andrew Zigler: It kind of, um, it, it decouples from the LLM provider from going to chat GPT or going to Claude directly. You know, it decouples those things like the tool [00:26:00] but also. Memory experience, personalization. That's an interesting customization tool. And also the one around security uh, or rather enforcing scopes because I know that's, uh. A paradigm, folks I've been grappling with. So, you know, maybe we'll follow up with you to get the, show the, a link to some of that for the show notes so folks can go check it out. Especially learn more about the, the, you know, the talk that you mentioned as well. Uh, I'm curious too, just kind of, um, on like a user level, you know, Brendan, are there MCP servers that you like to use that maybe you like to call out?
[00:26:29] Andrew Zigler: Or I'm curious, like how do MCPs fit into your, daily workflows?
[00:26:34] Brendan Irvine-Broque: I mean, I'm biased. I spend most of my waking hours probably, uh, here at CloudFlare. Uh, so I, I love. Working with our observability MCP server at CloudFlare, we've made a big bet there on just like, trying to make observability something that's first, first class on our platform and is built in. So I'm always using that to kind of debug things that I'm building or trying out on our platform for the first time.
[00:26:56] Brendan Irvine-Broque: I think that there's some really phenomenal [00:27:00] examples, uh, that were built, uh, on CloudFlare by the team over at at Block and Square. I remember seeing a demo when we did our demo day with them and they built this really cool thing where. You could, take a picture of a menu and then and analyze the menu, and it would go and take the, the, the actions and the tool calls to kind of add that to a storefront where you could.
[00:27:25] Brendan Irvine-Broque: Order online and see the menu. And like that became, that store kind of went from being in the physical world only to being online. And you think about how many steps there are in that process. But what's blown me away is that you can feed these tools, you know, higher order. Here's what I'm trying to do. Here's the project, here's, you know, what needs to happen, and they'll do the breakdown and all the kind of like,
[00:27:48] Brendan Irvine-Broque: boring administrative work and you end up in a situation where you, you feel very organized even if you're not an organized person.
[00:27:54] Andrew Zigler: Yeah. I really identify with that. I, and I like personally using the MCP tools where I can envision how they [00:28:00] can be used with a bunch of other tools. I, I typically don't pick up the tool. I don't see it as like, oh, this is like a baseline primitive for working on x, y, z, like, you know, maybe something that accesses my Slack messages might be something that's useful because I can use as a launching point for a lot of different workflows, projects and, and things like that.
[00:28:17] Andrew Zigler: So, and I, I, I love that you gave a shout out to Block. We love Block here at Dev Interrupted. Especially everything that they're working on with Goose, I've, I've done live streams with them as well with Goose and then trying out different MCP servers. So we, we love everything over at Block and, uh, highly agree that there's some really cool innovation happening out there and as we've kind of talked now, you know, this whole time about the MCP world and how it's grown within CloudFlare and Cloudflare's, really unique perspective on this whole problem, uh, we've kind of dug into why remote MCP is important, but is there anything top of mind that you wanna leave our audience with?
[00:28:52] Andrew Zigler: Because I think that there's going to be new revelations on the future about how this tool's really used and how things like [00:29:00] MCP can deliver even like, UI experiences into chat. Like there's this new thing it's relatively new, at least MCP ui. I'm not sure if you've, if you've heard of it, with being able to basically render iframes basically, you know, html fragments almost like a shadow dom within, uh, a chat, platform.
[00:29:19] Andrew Zigler: And so the idea of bringing the experiences of tools and other platforms, not just. Accessing the APIs through MCP, but actually bringing the experiences into the AI chats we're all using, that seems like a really cool future and I'm curious if you're, if you've seen that and what that looks like to you.
[00:29:36] Brendan Irvine-Broque: Yeah. Shout out to IDO and the team working on MCP UI. I remember when this was just like this. Early idea that a couple people were talking about, and that feels like yesterday.
[00:29:45] Andrew Zigler: know
[00:29:46] Brendan Irvine-Broque: I'm, I'm blown, blown away by how quickly people have been able to make this happen. I mean the, the, the mental model that I get excited about is that you know, if we really take seriously this premise that, your AI agent is [00:30:00] kind of a primary.
[00:30:00] Brendan Irvine-Broque: interface, not that like the web browser is going away or anything like that, but like so many people are turning to that first what, you know, what you might want to expose as a business. Back to those tools, you know, you want to have some degree of influence over how that's presented back to a person. And at the start when there was no kind of concept of, of UI, that a MCP server can, can return back.
[00:30:28] Brendan Irvine-Broque: It felt like, oh, like this is all kind of up to the agent to decide how this all gets rendered. You know, if I give you geolocation, will you render a map correctly? And what's exciting is the idea that there could be a future, at least to me, where, uh, you, you can go to these tools and get the benefits that come with that, but that the services that you're talking to can still have some influence over the shape and the style and the presentation of things 'cause that.
[00:30:54] Brendan Irvine-Broque: They're probably the best position to, to say how that information, that content should be [00:31:00] presented. But then that there, there, you create these, create these opportunities that I think a lot of people have wanted on the web for a really long time around being able to standardize. Interface and like design units.
[00:31:13] Brendan Irvine-Broque: D kind of design systems, right? Like why is it that iOS apps are often really nice to use is that there are these kind of UI patterns that are fairly hard to implement, like UI engineering's hard, but they're part of a system and you can kind of benefit from them for free. I do wonder a lot about if MCP UI is like a way into that for the web in a way that many people have tried in the past of web components and all the, the, the different kind of shots on goal that we've had.
[00:31:41] Brendan Irvine-Broque: Um, So I've had some great chats with IDO, uh, Michael Grinich, who's wonderful at WorkOS who worked on RADIX, uh, about this and excited for the future.
[00:31:49] Andrew Zigler: Yeah, we are too, and excited to keep chatting about things like this, and it's been really fun, you know, Brendan, to have you on the show and to get your perspective. Really appreciate you spending some time with us, uh, and lending your [00:32:00] perspective to our audience. And just before we wrap, is there anywhere that, uh, obviously everyone knows what CloudFlare is but is there anything in particular you want to, uh, direct folks attention to that's maybe top of mind
[00:32:11] Brendan Irvine-Broque: Yeah.
[00:32:11] Andrew Zigler: it
[00:32:12] Brendan Irvine-Broque: Go check. Out agents.cloudflare.com. We're doing a ton in this space of helping people build AI agents Check out the CloudFlare blog. We got a lot of interesting stuff coming. Love to, you know, reach out, meet different people. Use CloudFlare. Send me. Things that are good and bad.
[00:32:28] Brendan Irvine-Broque: Um, We always love to hear the feedback from everybody out there. We take our kind of role in the ecosystem pretty seriously. You can reach out to me at, uh, at Irvine broke. Uh, it's my last name and um, yeah. Thank you.
[00:32:39] Andrew Zigler: amazing, we'll, we'll make sure we include all of the. The links, the stuff that we talked about, like these blog posts and the talks and stuff, and in the show notes for today. And, and so to those listening, thanks for joining us here on Dev Interrupted. Uh, like Brendan said, please um, you know, reach out to us on LinkedIn, continue the conversation.
[00:32:55] Andrew Zigler: We want to hear from y'all about what you thought about today's conversation. What are you doing with [00:33:00] MCP servers. And until next time, we'll see you here on Dev Interrupted.
