The B2B Podcast Index
Index
All categories
MarketingSalesSaaSFinanceHROpsLeadershipCustomer SuccessAI & DataProductStartups & FoundersRevOpsEngineering & DevTools
MethodologySubmit
Best of:MarketingSalesSaaSFinanceHROpsLeadershipCustomer SuccessAI & DataProductStartups & FoundersRevOpsEngineering & DevTools
An independent project byFame
SearchBest episodesGuestsInsightsMethodologySubmit a podcast
Index/Ops/The Connectivity Cloud Podcast
The Connectivity Cloud Podcast artwork

Why the Next Big App Might Be Built by an AI Agent

The Connectivity Cloud Podcast · 2025-09-29 · 24 min

0:00--:--

Key moments - from our scoring

Substance score

37 / 100

Five dimensions, 20 points each

Insight Density7 / 20
Originality6 / 20
Guest Caliber11 / 20
Specificity & Evidence8 / 20
Conversational Craft5 / 20

Mark Dembo showcases Cloudflare's vision of removing friction from the developer experience, starting with a live demo of deploying a React website in under two minutes to 340 data centers worldwide without configuring instances, SSL certificates, or scaling. He then illustrates how GenAI tools like Claude can be extended into autonomous agents using the Model Context Protocol (MCP) - a new standard enabling AI systems to interact with external APIs and services. The demonstration shows an AI agent independently diagnosing errors in a URL shortener application by querying Cloudflare observability tools, analyzing logs and code, implementing fixes, and deploying to production with minimal human direction. Cloudflare positions itself as the infrastructure layer for agentic AI, hosting MCP servers for companies like PayPal, Stripe, and Wix, while emphasizing security and observability. The talk directly addresses developers and platform engineers building AI-augmented applications, particularly those seeking to accelerate deployment cycles and shift routine operational work to AI systems.

Key takeaways

  • →Cloudflare's developer platform enables shipping production-ready applications in minutes without manual infrastructure configuration, with automatic global distribution across 340 data centers and built-in security and DDoS protection.
  • →Agentic AI systems can autonomously diagnose production issues, make code fixes, and deploy changes by combining goals with access to system APIs through the Model Context Protocol (MCP) standard.
  • →The Model Context Protocol allows AI agents to interact with external services like Cloudflare's observability tools, enabling AI to independently gather context and make informed decisions without step-by-step human instruction.
  • →Cloudflare is strategically positioning its infrastructure as the deployment platform for MCP servers from major SaaS companies, giving enterprises a unified place to run, secure, observe, and control AI agents.
  • →Over 3 million developers now build on Cloudflare's platform, and the combination of rapid deployment with agentic AI automation significantly reduces time-to-market for both traditional applications and AI-augmented solutions.

In this episode

  1. 1Cloudflare's Vision for Developer Velocity
  2. 2Live Demo: Deploying a Website with Cloudflare CLI
  3. 3Introduction to Generative AI for Code Development
  4. 4What Are AI Agents: Autonomous Decision-Making Systems
  5. 5Live Demo: AI Agent Debugging URL Shortener Application
  6. 6Error Identification and Code Fixing with Claude
  7. 7Cloudflare's Strategy for Securing and Deploying AI Agents

Mentioned

CloudflareClaudeAnthropicPayPalStripeWixBitlyReactMark Dembo

Topics in this episode

Model Context Protocol (MCP)Claude (Anthropic)Observability and monitoringCloudflare WorkersGenAIAgentic AI agentsCloudflare Developer PlatformURL shortener applicationReact deploymentCloudflare MCP servers

Questions this episode answers

How long does it take to deploy a full-stack application to Cloudflare from scratch?

In the demo, a React website was deployed from initial scaffolding through DNS configuration and global distribution across 340 data centers in approximately 2 minutes, without manual configuration of instances, operating systems, SSL certificates, or multi-region setup.

What is the Model Context Protocol (MCP) and how does it enable AI agents?

MCP is a new standard that allows AI agents to interact autonomously with external systems and APIs. Cloudflare has built more than a dozen MCP servers so AI agents can query logs, retrieve configuration, read code, and trigger deployments without step-by-step human instruction.

Can AI agents deployed on Cloudflare autonomously fix production bugs?

Yes, as demonstrated with Claude using MCP, an AI agent can identify root causes by querying observability data, analyze code to understand errors, implement fixes, and deploy to production when given a high-level objective without detailed step-by-step instructions.

Which major companies are deploying AI agents and MCP servers on Cloudflare?

PayPal, Stripe, and Wix.com are among dozens of the biggest SaaS companies deploying their AI agents and MCP servers on Cloudflare's developer platform.

What infrastructure features does Cloudflare provide automatically without developer configuration?

Cloudflare automatically provides global CDN distribution, SSL/TLS certificates, DDoS protection, multi-region redundancy, auto-scaling, DNS configuration, and IAM - removing the need for developers to manually configure operating systems, instance sizes, or scaling policies.

What our scoring noted

Our reviewer’s read on each dimension, with quotes from the episode.

Insight Density

7 / 20

The episode is dominated by live product demos that consume roughly 70% of the runtime, leaving minimal space for substantive insights. When ideas are presented - like AI agents autonomously deciding next steps or MCP servers - they are explained superficially without deep exploration of trade-offs, limitations, or implementation challenges. The few claims made (e.g., 'never been easier to ship code') are assertions rather than evidence-backed insights.

What are AI agents? It's systems that can act autonomously. They make decisions and they adapt based on the environments that they're in. And all they need is goals.
It's never been easier than today to actually build something.

Originality

6 / 20

The episode rehearses well-worn talking points about AI-assisted development, serverless platforms, and global deployment - all standard fare in tech podcasts circa 2024. There is no contrarian thinking, first-principles analysis, or counterintuitive claim. The framing of AI agents as autonomous decision-makers is presented as novel but is now conventional. No fresh perspective on risks, economics, or team dynamics emerges.

We want to build a modern cloud platform that drives developer velocity, make you go fast.
What if the AI could actually do more than just the building?

Guest Caliber

11 / 20

The speaker (Mark Dembo, identified as making 'developers happy' at Cloudflare) is a practitioner with clear operational responsibility at a major infrastructure company. However, the episode is structured as a vendor product pitch rather than a genuine guest interview, limiting the depth of third-party perspective. The speaker's seniority is evident but the format undermines the value that might come from independent scrutiny or alternative viewpoints.

What's my job? Making developers happy. Really simple, straightforward, happy and successful.
As of end of last year, it's more than 3 million people developing on the Cloudflare developer platform.

Specificity & Evidence

8 / 20

Some concrete details appear: 3 million developers on Cloudflare platform, 340+ data centers, named companies (PayPal, Stripe, Wix), deployment timelines (120 seconds for initial setup). However, most claims lack quantification: no performance metrics for AI agents, no data on error reduction or business impact, no adoption rates for MCP servers, and no specifics on security or observability features promised. Demos show what tools do but not their effectiveness or cost implications.

As of end of last year, it's more than 3 million people developing on the Cloudflare developer platform.
That was roughly 120 seconds.

Conversational Craft

5 / 20

This is a stage presentation, not a dialogue. There are no follow-up questions, no genuine disagreement, and no push-back on claims. The host (if present) does not challenge the speaker on limitations, risks, or trade-offs. Technical obstacles during demos (GPU overload, demo failures) are deflected with humor rather than explored. The structure is entirely one-directional: assertion → demo → assertion. No intellectual tension exists.

Demo time. Those slides are my favorite because I can just copy paste them a bunch of times.
I really hope that this works.

Conversation analysis

Computed from the transcript - who did the talking, and the words that came up most.

Most-used words

cloudflare25code15check14website12last12developer12didn12agents11errors11today10build9deploy9show8cool8works8seconds8

Episode notes

In this episode of The Connectivity Cloud Podcast with Cloudflare, host Mark Dembo shows what it really means to build without friction, from spinning up and deploying a live website in 120 seconds to transforming it with GenAI and then debugging it using agentic AI. What You’ll Learn How to deploy global web applications in minutes without managing infrastructure, certificates, or scaling The practical benefits of Cloudflare's automated multi-region deployment across 340+ data centers Why agentic AI represents the next evolution in development How AI agents can independently diagnose and fix production issues by interfacing with deployment platforms Mark Dembo serves as a Senior Manager at Cloudflare, where he focuses on "making developers happy and successful." With vast expertise in cloud platforms and developer tooling, he leads initiatives to enhance developer velocity and streamline application deployment processes. At Cloudflare, Mark has been vital in building a platform that serves over 3 million developers worldwide, focusing on removing friction from the development process and enabling rapid deployment at enterprise scale.

Full transcript

24 min

Transcribed and scored by The B2B Podcast Index.

Speaker A: What's my job? Making developers happy. Delete everything and then. Hello, Munich. I know. That is insane. This will be the most exciting website we'll have ever seen. As of end of last year, it's more than 3 million people developing on the Cloudflare developer platform. And I have one last question for you. So what will you build next? Welcome to the Connectivity Cloud podcast, the podcast that provides expert insights into the cloud and IT landscape. Hi, I'm Mark Dembo. And each month we'll explore key topics like scaling secure infrastructure, tackling emerging risks and staying ahead of the latest trends. Whether you're managing multi vendor environments or navigating cloud modernization, this is the show for you, delivering practical advice for today's decision makers. For people who don't know me, I'm Mark. Nice meeting you. Uh, what's my job? Making developers happy. Really simple, straightforward, happy and successful. What's Cloudflare's job? Enable developers to build good things fast. That's all I want you to know from this talk. If you want to go take a phone call, that's it. Otherwise I have more to show you. But that's the core message. Okay. That was not the approved marketing message. That is. So let's do that slide as well. Okay. Cloudflare's vision for developers. Really simple. We want to build a modern cloud platform that drives developer velocity, make you go fast, whether that's go to market experiments, shipping, modernizing your application, stack, all these kind of things and drive better end user experiences. How do we do that? We want to remove friction. We have all the Lego bricks that you need to actually build stuff. We do that at enterprise scale and also in an AI ready world. Sounds great, right? Everyone convinced? Not yet. Okay, cool. Let's do a demo then. Okay. And I'm already praying to the demo gods. We are doing lots of demos today. So I told you about, we want to go fast. What we're doing right now is building a website because I feel like that's the right thing to do on stage. What could go wrong? And what I'm doing here is I went to the Cloudflare documentation so I cheated a little bit. We are like bringing up a react website and I'm using the Cloudflare CLI for that. It's asking me a few questions. So, uh, let's quickly maximize that. Yeah. And I want to proceed with the stuff. So that's awesome. And I hope the WI fi works and I'm not stalling whatsoever for like 20 seconds and it's making progress. Step two out of three. That's amazing. And within a few seconds we'll have like a template, right? Not like a major thing, but it's like a hello world of this project life. And because I'm lazy, I click yes for everything. And right now what we're doing is we have set this up locally, let's deploy this to Cloudflare. So I select my account and we're building this. And right now, any second now, we're actually right now uploading to Cloudflare servers. So that's cool. It is uploaded and what we're now doing is setting DNS entries. Not fun. So that might take a, uh, brief second. And once that's done, we have a website we can visit. Let's check that out. I just deployed that. So that was roughly 120 seconds. That's nice. However, that's just the starter template, actually. I want to show you how much of a great coder I am. So let's go into code editor. And what I'll do is firstly zoom in so you can actually read stuff. And then let's change a little bit what the content here is. And for everyone who can code, you'll see my elite coding skills in just a second. Delete everything. And then, hello, Munich. I know. That is insane. This will be the most exciting website we'll have ever seen. We hit save. Let's check in local development whether that looks good. That looks great. I actually want one more thing because I'm a friendly person, at least I believe so let's add an hour shot. You know, there goes. Always works with the audience. I heard you, I heard you. Works every single time. You're so predictable. Okay, let's hit deploy right there. And you'll see. Okay, it wants to update. I don't want to run an update now worth timing error. But you'll see. I just did a change to my website also in just a few seconds, right? So it's uploading and it will say, hey, yeah, it's working. Yada, yada, yada. And then if I refresh and I think I might be too impatient here. Oh, that's my local website that is now deployed in 330 data centers all around the world on Cloudflare. So that's awesome. So what did we just do? Well, I guess we shipped an application. Exciting. I know, but what didn't we do? We didn't scaffold a project manually. We did not configure instance sizes. We did not select and manage operating system RunTimes deal with SSL, CERTs, DNS entries, multi region setup, multi instance setup for availability. So multi region and yeah, availability zone. We didn't configure scaling, we didn't put a CDN in front of that though. This app that I've put up on there and I admit it, it's not too exciting today. You could hit it with a million rps, uh, and it would just work. This website is being shipped to 340 data centers all around the world. So if one data center goes down for whatever reason, there's 339, I guess to go. You didn't do any of that. We did that. So that's what we're talking about when we say we want to remove friction and we want developers to go fast. So what did we actually do? We shipped the app with great user experience. By default it's highly available, it's globally distributed, it's close to end users, it's cache where possible, it's secure, you have um, like SSL, all this kind of stuff, DDoS protection, all is there. We didn't do any of that, we didn't do any IAM config, all this kind of stuff. It's all done for us. And that's actually convinced million of developers. As of last year, end of year last year, it's 3. More than 3 million people developing on the Cloudflare developer platform. That's pretty cool. That excites me by the way, because my job is to make devs happy. More people I can make happy. So how do we beat that? You might have heard of this, there's a thing called cool. So let me show you what that does. So we'll take our brilliant website and I will go again. Oh, I have the code editor already open and it makes me look like a really bad developer. But let's do something. So let's say, hey, I need your help, please, we want to be polite. Change this website to create a quiz on the Cloudflare developer platform. Make it entry level, fun and fun. And you'll see now why I'm elite AI specialist. Because I have a lot of tricks up my sleeve and one of them is like make it look awesome. Okay? And that is what Genai brought to the table, right? It is us telling the AI hey, that's what we want to do. It's working. It's giving me the code output and that looks great. I guess not entirely sure, but let's do it. And it's telling me, okay, let's apply this so it's now applying all this kind of stuff. And I'll just say that looks great, I guess, because I don't have time for code review right now. And there is also, like, to make it for a more fun look, I should change this. I'm very tempted to change this. And once we got that, I will say save that as well. And because I'm not a guy who has a lot of time today, I'll just approach the Internet. Uh oh. Ooh, what did I do? One second. Today, REACT is not. AI and I are friends, but sometimes we have our better times than worse times. I really hope that today I can show how much of friends we are. So within a few seconds, we should see that. And I want to do one more thing actually, while I'm waiting. I'm a little bit of a multitasker. I also want to show. Hey, right now it has this. I don't want to say ugly URL, but how about we give it a different URL and I give it a pattern. I call this Munich CF Quiz. And I have a domain which is a little bit. If you get the joke, you get your joke. But it's like I'm a developer and not a single bug.com is. I felt what a good domain name looks like. I say, that's a custom domain. And once I deploy again, not only will have this, like, work as a dev domain, but also it will be exposed on my custom host name. And that's what I just built in, like, roughly, I don't know, two minutes. What is Kafka workers? I should know this. So, uh, let's try this together. I actually don't want to do this. I might embarrass myself. So let's move on. Right, okay, so that's what changed. Genai demo time. Now the question is, where do we go from here? Because I still have time on my presentation. I actually have plenty of time. What if the AI could actually do more than just the building? Like right now? If you look at what a developer does, we design stuff, we do specs, at some point, we translate that into code, we deploy. But using modern team setups, you also operate your solution in production, you monitor it, you look at what's working, what's not working, you improve and iterate based on that. And you do do bug fixing, right? Well, 20, 24 and onward, we have this term called agentic AI that sounds maybe a little bit abstract. Let me try to clarify what's happening there. What are AI agents? It's systems that can act autonomously. They make decisions and they adapt based on the environments that they're in. And all they need is goals. So it's a little bit more abstract and a little bit more powerful than what we just saw. And I want to show you how that works. Demo time. Those slides are my favorite because I can just copy paste them a bunch of times. But then being on stage is a little bit more stressful than if you just did the slides. So maybe I should be less lazy in preparation. So I built this awesome website today. It's called Short. Not a single bug. Com. Don't use it for your production environments. Don't rely on it, because I might delete it on or break it. But you get the idea. What does it do? It's a really simple thing. It's like, you know how sometimes you have, like, these URLs, which are, like, really long, and you, like, just want to, like, shorten it? Guess what? You enter URL here, click on shorten URL. And it's wild. You can just like, click on it and you get to the actual link. I mean, like, basically Bitly or whatever. I built this, took me like 10 minutes. I felt like, that's a good app that everyone understands. Nothing crazy, right? Okay. And the guy who built this might have not done an awesome job of handling all user inputs. So let's check what happens if I hit, oh, a URL that's not a URL. Oh, we have an error. That's not great. What about, I don't know, me adding some numbers or starting with special characters? Whoever coded that? Not entirely great, right? So something is wrong in our code. And I built this also on the Cloudflare developer platform. So let's check out what we see there. So I'm here on the Cloudflare dashboard, I'll maximize this and you'll see there's an application that's called Notmenic. That's what we just deployed, by the way. That's what we built is also on my dashboard, of course. And then there's this URL shortener. Let's check what happened the last five minutes. And you will see, oh, a few errors here. So we have a few successful requests. That what happened when I showed you that it's actually working. But we also got some user errors, like, really easy. Right. What I would typically do now as a developer is like, I would go to the logs, I would check the log messages, try to make sense of them. I get a stack trace if I'm lucky. If I didn't swallow the errors or like add console log statements whatsoever, then I go to my code, I figure out and try to correlate and see what's going on. Well, I told you what if we could take a different approach? So let's do just that. And so I brought another of my AI friends. He's called Claude. And Claude. Claude is a tool that runs in your command line. It's just basically a tool from Anthropic and it's for developers. Let me see whether Claude can fix this. Some of my users running into weird issues with the application. Can you check errors for the last 10 minutes and identify the root cause would be awesome. Be friendly. You never know what they're going to do next year. Okay, so what I just did there, not technical whatsoever. I describe what I want to do. I am telling it. Okay, great. And right now, this is my worst nightmare. I really hope that this works. What I'm doing is becoming really, really popular. So the people from Anthropic that are powering this LLM, they don't have enough servers. But on 5th, uh, attempt it actually got working. So just ignore when that happened, it didn't work. Shit. We'll see in a few seconds. But I really hope that I will get it to work. Um, because when it works, it's amazing. It's just too many people here using this right now. It's blowing up like crazy. I might have to retry. It doesn't matter. Like I can talk for that period. And you love hearing me talk, so. So this, this will be great. But, but long story short, I just give you some objectives, right? We talked about this. What is an AI agentic system? Right? It's okay. I might have to just restart demo. Gods, I need you. Okay. I'm just cheating by repeating my message from last time really quickly and still overload it. I hope that we'll get this work, but what did I tell you? I give this system a goal. I'm telling it, hey, please identify this and it comes up with a game plan on how to do it itself. Ideally, if this works right? So cool. What it just did is here it came out with a, uh, to do list. So what will it do? It said, okay, I'll first check what happened the last 10 minutes. Then I'll try to analyze the errors and check like in general, whether things are healthy. So it came out with like a game plan, what it wants to do and what it's doing right now based on that game plan. It's Deciding itself what it's going to do next. So what it did is it read my configuration here, it's getting the current time, and based on that it says, okay, cool. Oh, it wants to do something differently. Let me guide a little bit because my demos don't want to work, use the MCP server instead. Not happy with the demo gods today, but that's okay. So it comes up with a game plan and ideally you roughly know what you want to do, uh, you want it to do. And now it's overloaded again, but we have still plenty of time. It will work, I'm not worried. And literally came up with those to dos. I didn't tell it step by step what to do, and it's executing stuff right there. And what it did previously, it said, hey, most likely to figure out what's going on with the error logs, I need to figure out like what is actually configured. So it read my configuration file, then it said, hey, if I want to get the last 10 minutes, I need to know what time it is. So it ran a command to get the current time. And what it's doing right now is it's running a Cloudflare MCP service. So while this is running, what it's doing is it is actually decided, with my help, a little nudge, to actually use the Cloudflare observability mcp. MCP is a new standard or protocol that allows AI agent to interact with other systems out there. Cloudflare has built more than a dozen of these, so you can check for instance, your logs of your work applications using just agentic AI. So what this tool is trying to do, if it can get a hold of one of those overloaded GPUs in the cloud, is actually to call Cloudflare and interact with it. And I didn't do any of that. Ideally, by the end of this, once I hope this works, it will try to adapt its game plan based on what it's seeing in those logs. So let's see whether I can get this to work. It would be such a pity if it didn't because it worked so well. But once the US comes online, I kid you not, this, this is a competitive advantage of living in Europe. You get GPUs in the morning, I kid you not, is exactly what's going on. So what it's doing is it read from my configuration, hey, this is my account id. So it's setting that and yeah, saying, okay, cool, I'm doing this, I'm setting my account, and again, it's Running out of GPU capacity. This would be so smooth without. Let's see how well this is going. Don't we worry about. So, yeah, and it says, I want to get all your workers that you have deployed. So I'm also doing that. I'm saying, yep, that sounds like a good game plan right now, by the way, this is human in the loop mode because it's really nice for my demo. You could also just say, AI, do your thing, grab a coffee, come back and we'll just do stuff. Personally, I'm not at that trust level yet, but some people are maybe more YOLO than I am, so have fun. And what it's doing right now is saying, hey, I want to get, for a given worker, this URL worker here, I want to get, for the last 10 minutes, I want to get the error logs. So it's actually based on my prompt, figure out what to do. It got the right worker, it got the right time, it reused the right tool. And it's literally right now independently talking to Cloudflare to try to figure out what's going on. So ideally it will come up with a solution, but it might not be because we don't have time to wait on GPUs. But that is the powerful thing, right? This is the agentic way of doing things that it will basically do decisions based entirely on what it sees in previous, uh, steps. Right? It chooses which worker to get logs from based on the configuration it read itself. Right. Now it's decided, hey, I want to read the code here, actually, because that might help me to debug the issue and if I actually read the code for context. So that's what it's trying to do there as well. And then it says, okay, great, I want to also check the KB Store health and connectivity. Not sure what that's all about, but if the AI wants to do that, why not, right? Might be a good assumption because that solution that I'm using under the store uses the Cloudflare KB service. So if I want to check whether the service why it's throwing errors, that is a good thing to check, right? It might not be the issue here because I might have, like gone for something else, but that's literally what's happening, right? So it found it and it says, okay, you know what the root cause is? You're throwing invalid URL string on line eight. Well, we could change that basically, and you should add validation before trying to parse it. So it has found the root cause without me doing anything. I'll Tell it, can you please fix it again? You know me, I don't like to do work myself. I like to talk about the work that someone else is doing. I'm a manager now, so that's what's happening. And literally that's exactly what it did. It says, okay, cool, let's actually fix it. And it came to me and said, uh, oh, you know what? Instead of fixing this, let's actually check whether it's working. And if not, let's handle the error gracefully and say, it must be a string. It should be in the right format. Do it. And I want to. Do you want to edit this file? Yes. Hell yeah, I want to do that. So that's what it's doing right now. And just imagine in a few seconds it will do that. And it already said, I want to deploy this fixed to production. So it's thinking ahead. Let's see whether it knows how to do that. Because that would have been my next step to say, hey, please deploy it. Ready to run. Yes, deploy it. And just like that, it. In a few seconds, we'll see. And, um, we'll see that in our logs, basically that, hey, it will not be able to parse those URLs, because, hey, they are still not good URLs. But what the LLM did, just to recap and just for you to comprehend how freaking crazy that is, I gave it a task with an objective, Please fix it. What it's done is like, interact with another system, which is Cloudflare. It interacted with my file system, my code base. It understood the error, it iterated each approach. It came up with a fix. It deployed this like. Or is deploying it to production. Basically got an output here, and basically it's done. And I just told it, hey, please fix and deploy. Thank you. And it says, deployed, it's done, go ahead. And it's saying, okay, now instead of getting, like, throwing errors. And you'll not see this in the front end, but we'll see it here. So let's monitor this. In the past it was saying, like, error is 3 and all these kind of things. Well, if we now go to the URL show and I give it something wrong, it will still tell me, hey, it doesn't work, that's fine. But what it will do is if we look here in our errors, those errors are handled with a nice API response, which gives the developer, if we just look at this like, it really, like, it gives you an idea of what's actually going on. Before that, when there was an internal server error and now it's telling us as a response, yeah, the URL is not in the valid format, you idiot. That's basically what it's telling the developer. And if we refresh here. Hey, you see, hey, yes, I see requests, but I don't see errors because our uh, code has improved that we actually handle those kind of errors. Okay, so what does that mean? We want Cloudflare to be the best place to deploy AI agents, secure AI agents, connect with AI agents, observe AI agents, control AI agents and build with AI agents. I've just showed you the last bit. I showed you built with AI agents and I've showed you connect with AI agents. Cloudflare, uh, we really, really, really investing into being the home for agentic AI. So not only do we have a lot of like MCP service so that you, you can build using agentic AI, that MCP server, guess where that's deployed? Also on the Cloudflare developer platform. And not only we are deploying those MCCP servers for ourselves. Some of the biggest brands in tech think like PayPal, Stripe, Wix.com, i don't know who else like we have dozens of the biggest SaaS companies on the planet that deployed their AI agents, their MCP servers on Cloudflare. If you know Cloudflare, we always think about security and this standard of MCP is evolving really really quickly and it's tough like to get the security right. So we are thinking about that. I cannot talk too much about it but read our uh, blocks and follow and you'll see more about securing, observing and controlling MCP servers. So that's the future of agentic AI and building with that. And I have one last question for you. So what will you build next? Because it has been never easier than today to actually build something. I've seen non technical people build like a website for their baby shower where people could select like which presents they're going to bring and whatever and they connected it to database and they don't know anything about coding whatsoever. I've seen organizations ship code faster and refactoring code, moving between providers, roof vendor lock in because they could ask like AI agents to actually do this and they move so much faster. So my question is it's 2025, it's never been easier to ship code. What are you building next? And are you building it on Cloudflare? Thank you so much. Thank you for tuning in to the Connectivity Cloud podcast. If you found today's episode valuable, be sure to subscribe so you won't miss future updates. Stay ahead of the curve, stay connected, and stay secure. As always with cloudfl,

Related episodes across the Index

Other episodes covering the same guests and topics, from across The B2B Podcast Index.

  • Telstra’s Time Sync Outage, DoorDash’s 1.5M RPS Cache, Stateless MCP, GitHub and PyPI Supply Chain Delays, and Why the Quietest Infrastructure Often Has the Biggest Blast RadiusShip It Weekly · on Model Context Protocol (MCP)97 / 100
  • The Benchmark With No Instructions - ARC-AGI-3 (winning team!)Machine Learning Street Talk · on Claude (Anthropic)85 / 100
  • The End of One Model to Rule Them All: Why Enterprise AI Is Going Small, Specialized, and Multi-ModelDisambiguation · on Claude (Anthropic)85 / 100
  • 304: Boom, bust, or bubble? Rock Health weighs in on digital health funding in 2026Radio Advisory · on Claude (Anthropic)84 / 100
  • AI for Engineering Is Leaving the Demo PhaseAI Across The Product Lifecycle Podcast · on Claude (Anthropic)83 / 100
  • Late checkouts, AI agents and the future of guest communication with Cole Rubin of ConduitMatt Talks Hospitality: Real conversations for innovative hoteliers · on Model Context Protocol (MCP)83 / 100

More from The Connectivity Cloud Podcast

All episodes →
  • Cybersecurity Predictions 2026: What Security Leaders Learned in 202566 / 100
  • Inside Cybersecurity's New Arms Race with Vladimir Krupnov and Blake Darché74 / 100
  • How This Bank CISO Can Redeploy His Infrastructure in Minutes61 / 100
  • How Retailers Are Fighting Back Against Fraud with Andy Dean and Christian Reilly81 / 100
  • How to Build a Security Team That Actually Enables the Business with Olivier Busolini
Explore the best B2B Ops podcasts →
All The Connectivity Cloud Podcast episodes →