Request // Response

Smithery Co-founders | Henry Mao (fmr Jenni AI) & Anirudh Kamath | Building AI-native services with MCP, developer distribution challenges & the future of agent tooling

In this episode of Request Response, I sit down with Henry and Anirudh, co-founders of Smithery, the easiest place to build, deploy, and discover MCP servers. We dive into how Henry and Anirudh went from building custom chatbots for specific use cases to creating infrastructure that makes MCP development accessible to any developer. We explore the fascinating world of Model Context Protocol—from its potential to revolutionize how AI agents interact with external systems to the practical challenges of building reliable, production-ready MCP servers that enterprises actually trust. Henry and Anirudh share their vision for making MCP as ubiquitous as REST APIs, and we discuss how the ecosystem is evolving from experimental side projects to critical business infrastructure. If you're working with AI agents, building integrations, or are curious about the next wave of API-like protocols, this episode offers a ground-floor perspective on technology that could reshape how we think about AI-to-system communication. Quotes from the Podcast Why MCP has gained adoption "MCP has won developer mindshare because it's really striving to be an open protocol. There were previous attempts to create AI app stores like the GPT store, but those approaches were more of a walled garden where you need permission from the company to be part of the ecosystem. MCP is different - anybody can implement it and your agent can work with it without needing permission from Anthropic. They're creating their own governance and making this more like HTTP or an open standard that multiple companies are contributing to. It's made for the developer community." The value of specialized AI services and agent orchestration "I think there is something to be said about specialized AI services. Like I was talking to someone at DeepMind who was of the opinion that building services itself is stupid because the models will get so good that you could just say 'build me Salesforce' and the model will do everything for you. But even as a human who is generally intelligent, I can set up my own software stack on EC2 and build my own Postgres database - I just choose not to because it's a giant headache. Building software has never been the hassle, it's been maintaining that software. What you're seeing is more specialized AI services, like V0 for front-end development versus using Cursor which is general purpose. I think it's agents all the way down. What we're trying to build towards is agents getting really good at handing off tasks and isolating tasks - so you can tell your agent 'build me Salesforce' with a $30 budget, and that agent should distribute the budget across front-end, back-end, and infrastructure, making intelligent decisions about which specialized services to use for each component." MCP reputation and discovery "Something interesting about MCP itself is the client doesn't actually know what servers it's connected to - it only has access to all the tools at once. If you have internet search and another tool called 'best really good amazing internet search', what does it choose between? There's no reputation score assigned, and it doesn't even know those tools could be from two different servers. What you could do with an auto-router is basically say 'this server has a bad reputation score' or 'this tool call didn't work on this server' so we're going to ignore that server altogether. There could be another layer to help with reputations."

Scott Dietzen (CEO, Augment Code) | Context engineering & enterprise-scale AI coding

In this episode of Request Response, I sit down with Scott Dietzen, CEO of Augment Code and former CEO of Pure Storage. We dive into why context selection has become more critical than prompt engineering, and how his team solved the fundamental challenge of giving AI agents just the right amount of codebase context to be effective without being overwhelmed. If you're working with large codebases, building developer tools, or wondering how AI coding assistance scales beyond startup-sized projects, this episode is a must listen. Show Notes [00:00:53] – From machine learning PhD to startup CEO: Scott's journey to Augment Code [00:02:05] – Anti-vibe coding: tackling enterprise-scale codebases with millions of lines [00:02:45] – Enterprise customers and graduating from Cursor to Augment [00:04:38] – The context problem: why LLMs struggle with massive codebases [00:05:39] – From prompt engineering to context engineering as the new bottleneck [00:06:35] – MCP adoption and helping ISVs package documentation for AI [00:07:22] – The risk of getting lazy with API design in an AI world [00:08:35] – Metaprogramming with agents and the importance of code review [00:09:22] – AI-generated testing improving coverage and unlocking legacy codebases [00:10:27] – The $2.5 trillion software failure problem and the promise of AI [00:11:22] – Backlog zero: the holy grail for developers [00:12:02] – Lessons from distributed systems: simplicity and reliability in commercial software [00:12:48] – Engineers as tech leads for agents: human judgment still essential [00:14:02] – Augment's differentiators: context engine, security, and IDE compatibility

AI-friendly API design buy vs. build, and scaling DevEx | Tom Hacohen (CEO & Founder, Svix)

In this episode of Request Response, I sat down with Tom Hacohen, founder and CEO of Swix, the Webhooks-as-a-Service platform. Tom shares the origin story of Swix—which started as a side project to escape the pain of webhook maintenance—and how it’s grown into essential infrastructure for API-first companies. We dive into why great developer experience means solving actual problems and how AI will reshape API design—from brittle CRUD endpoints to high-level abstractions agents can reason over. If you're building APIs, AI infrastructure, or just tired of rewriting webhook engines, this episode is a clear-eyed look at where platform engineering is heading—and what it means to ship something developers actually love. Show Notes [01:18] – The founding story of Swix and the underestimated complexity of webhooks  [04:24] – Prioritizing developer experience over implementation details [05:42] – Early solutions: replacing homegrown webhook infrastructures  [07:15] – Build vs. buy: the case for buying webhook infrastructure  [09:03] – The cost of maintaining v1 and the pain of v2 and beyond  [10:25] – Why companies rewrite their webhook systems multiple times [12:32] – AI’s limited current impact on build vs. buy decisions [14:24] – The human side of solution engineering and customer acceleration [16:29] – Tool calling with LLMs and the rise of abstracted APIs [18:19] – AI favoring flexible, NLP-friendly endpoints over brittle CRUD [20:07] – Toward workflow-aware API responses for agents [22:01] – Meta MCPs and hierarchical API discovery [23:08] – Swix’s role as AI infrastructure and aligning with AI-oriented standards [25:56] – What great developer experience really means Additional Quotes from the Podcast "No one cares about webhooks. They don't care about your SDKs either. They don't care about any of that. What they care about is either not having to build it or the value that their customers get." "You can build anything... The real question is: is it worth your time and effort? Is it worth re-learning? Because honestly, writing the code isn’t the hard part—learning all the complexities and edge cases is." "People often get developer experience wrong—they reduce it to 'good APIs and SDKs' or 'nice docs.' But it’s more than that. It’s about the entire experience—from onboarding to support to observability. Like UI design isn’t just colors, DX isn’t just code. Good DX is making the right thing easy and the wrong thing hard."

AI's impact on API integration patterns and building frictionless developer experiences | Anuj Jhunjhunwala (Director of Product, Merge)

In this episode of Request // Response, I sit down with Anuj Jhunjhunwala, who leads product and design at Merge. We talk about how Merge is transforming integration pain into product velocity with their unified API approach. Anuj shares how his team helps developers avoid the complexity of maintaining dozens of APIs, and we explore the rising strategic importance of integrations in AI-driven products. We also dive into what great developer experience really looks like and how AI is reshaping expectations around API design and usability. Show Notes [02:30] – The pain of building and maintaining custom integrations [03:45] – Merge as “Plaid for B2B software” and the magic moment for devs [07:30] – AI making integrations table stakes; the three types of data [08:30] – Why proprietary data is the key to differentiation in AI [09:45] – AI and the future of APIs: personalization and intuitive design [11:30] – DX vs AX: API design for devs and for AI [12:00] – How AI product patterns are changing API requirements [13:00] – Richer queries, delta endpoints, and evolving API design [14:00] – The rising importance of fine-grained permissions [15:00] – HubSpot, search endpoints, and future-facing API choices [16:00] – Delta endpoints explained and why they’re valuable for LLMs [17:00] – Principles of great developer experience: predictability and frictionlessness [19:00] – Where to learn more about Merge Additional Quotes from the Podcast API Integrations are Table Stakes for AI "The biggest trend is that it's just becoming tables stakes, API integrations. We talk to a lot of AI companies. We get a lot of interest in AI companies, and there's really like three types of data. That is helpful when you're building an LLM, right? There's the public data that exists out there that's scraped from the internet and is publicly accessible. There's synthetic data, which is, you know, obviously it produced itself by an algorithm or by an LLM and can help you validate and test out edge cases. And then there's proprietary data, right? So it's the data that belongs to your customer. And the first two things there, the public and the synthetic data. Are accessible to basically anybody. The proprietary data is a data that's specific to you and makes you different, and it's a thing that no one else can get access to because it literally just belongs to you or your customer. That I think, is the magic of the integration, is that the integrations pull in that third bucket and like they can make your product different, so you're leaving money on the table if you don't have an integration strategy because you're not thinking about that third bucket, which actually makes you different." AI's Impact on API Design and Documentation "APIs in general just have to be extremely intuitive and easy to get started, right? So like AI, I think, spoils us at a very high level into expecting a lot of personalization, right? When you go to ChatGPT and you ask it a question, it can feel like a very pointed personal response to what you just said. It's a unique answer in many cases, like the exact question you have. Separately, but also relatedly, no one reads docs, right? No one reads anything. Um, they, you know, they just, uh, I'm at the point where if I use a third party tool or if I am figuring out how to troubleshoot something at home, the first thing I do is open up ChatGPT and like take a picture of it and say, how do I use this thing? And so I think people are using AI to personalize the type of responses they get. They expect low friction in all these encounters." What is Great Developer Experience?  "I think ultimately what this boils down to is that great DevEx is reducing friction to the point where your team can spend its time on other high-leverage things that you'd rather spend your time on, right? That's ultimately what this is—providing leverage. It's multiplying your team. It's a force multiplier, and I think that's super powerful just because out of the box you click a few buttons on a website and it can do that for you, which is a really powerful experience in my mind."

Overcoming barriers to mobile dev, smoothing DevEx, and AI-driven coding shifts | Charlie Cheever (CEO, Expo)

In this episode of Request // Response, I sit down with Charlie Cheever, CEO of Expo and co-founder of Quora, to unpack the evolution of mobile app development and how developer experience is adapting in an AI-assisted world.  Charlie shares stories from scaling Quora's mobile presence, his frustrations with App Store complexity, and how Expo is aiming to make app development as smooth as deploying a website. From React Native to GraphQL to vibe coding, Charlie breaks down the current gaps in frontend-backend integration and offers a wishlist for what a truly great developer experience could look like—particularly in a world where more non-developers are writing production code. Whether you're building mobile apps, exploring streaming APIs, or thinking deeply about DX tooling, this is a must-listen. Show Notes [00:00:00] Introduction * Mobile app development, developer experience, Quora, and Expo. [00:01:00] The Challenge of Mobile vs. Web Development * Why mobile app development remains more complex than web. * Charlie shares the motivation behind starting Expo. [00:02:00] Quora’s Mobile Journey * How building native iOS and Android apps took longer than launching the original website. * Technical and organizational challenges of scaling mobile. [00:03:00] Building Expo and React Native’s Rise * The inspiration behind Expo: make native development as simple as the web. * React Native’s explosive popularity and Expo’s role in democratizing it. * Helping teams without Meta-scale infrastructure. [00:04:00] Making React Native Work for Everyone * How Expo bridges gaps in native app development. * The importance of tools like Expo and Vercel in modern developer workflows. [00:05:00] Is GraphQL Still the Right Abstraction? * Charlie reflects on GraphQL’s evolution and current limitations. * Comparing it to frontend framework history (Rails → React). * The need for more end-to-end client/server coherence. [00:06:00] Balancing State, Performance, and Cost * Complexity in backend design due to varied use cases. * The challenge of making highly reactive systems affordable and performant. [00:07:00] Streaming APIs, SSE, and DX Gaps * How the rise of real-time UX (e.g., chat apps) is stressing API design. * No streaming standard in OpenAPI/GraphQL specs yet. * The tooling gap for reactive and streaming data patterns. [00:08:00] Wishlist for the Future of API Dev * Charlie’s top priority: cost-efficiency in scalable systems. * The hidden costs of AI-generated or streaming-heavy applications. * The importance of default patterns that are economical by design. [00:09:00] AI, Prompt-to-App, and Developer Onboarding * The influx of non-developers using Expo via AI-powered tools. * Reworking product terminology and UX for AI-native builders. * The shift from code complexity to deployment complexity. [00:10:00] Vibe Coding and System Architecture Risks * The risk of vibe coding critical backend interactions. * Why grounding AI-generated work in opinionated, proven patterns matters. [00:11:00] Scaffolding and Guardrails for AI-Driven Dev * The role of SSA-style rules and architectural constraints. * Importance of enforcing standards even in semi-AI-written codebases. [00:12:00] A Hybrid Dev World * Current state: most repos are still human-written with AI augmentation. * The mismatch when AI-generated code doesn’t follow team conventions. [00:13:00] Expo’s Strategy for the AI Era * Adapting Expo for AI-assisted builders. * Making deploy tooling more accessible as code creation gets easier. * Value shift from writing code to shipping product. [00:14:00] The Magic of Going Live * How products like Vercel V0 and Expo reduce the friction of launch. * Sharing live URLs as the new MVP moment. [00:15:00] Great Developer Experiences: Inspirations * Charlie’s favorite DX examples: PHP, early Rails, undo features. * Simplicity, feedback loops, and emotional empowerment as DX pillars. [00:16:00] Undo and Developer Safety * Git, Neon, and the psychological benefit of reversible workflows. * Why safety enables faster and more creative software building. Additional Quotes from the Podcast How AI is Changing the Way Users Build Apps "A huge number of the people sort of signing up for Expo accounts and stuff now are using AI sort of prompt to app tools. So we're having to build a whole new set of products or at least change some of our products 'cause some of the terminology that we use is, you know, tailored to developers and like developers know what this means or that means. And a lot of times these people are like, "I am using Expo because I saw the name come up a bunch of times as I was, you know, prompting. But I don't really know what it is and I don't know why I'm here. And what are you gonna help me do and why do I need to?" It was always like, "Hey, like there's always people who just have ideas. How do we make those come to life as quickly as possible?" And so like for a long time it just sort of felt like, well there's this whole hairy problem of doing the writing, the code of software development, you know, writing React code, writing backend code or whatever. And like all of a sudden that's getting way, way easier and going way, way faster." The Magic Moment That Defines Great DX "I think that the biggest thing that like you can give in sort of a Dev X experience is if you take somebody who like thinks they can't quite do something and that's got, something's gonna be hard, and then all of a sudden it just ends up easier than they thought. And they're just sort of on a smooth path to something happening. I remember watching videos of people competing like in the sort of mid-2000s where, and I never even used Rails, but I just saw these where it's "make a blog in 12 minutes with Ruby on Rails." And then somebody else had put out a video that was like, make a blog in, you know, eight minutes and then seven minutes and then six minutes and just sort of like competing to get that as smooth and streamlined as easy as possible was I think pretty amazing."