S-Tier MCP Servers for Developers  ## Metadata - Author: [[Syntax]] - Full Title: S-Tier MCP Servers for Developers - Category: #articles - Document Note: --- tags: [MCP, Developer Tools, AI Agents] author: [MISSING] source: Syntax (YouTube) reference: [https://www.youtube.com/watch?v=5az_4IwTwAE] date created: {{date:YYYY-MM-DD}} {{time:HH:mm}} --- # 🏷️ Title: S-Tier MCP Servers for Developers ## 🧭 Table of Contents - [📝 Summary](#-summary) - [📖 Main Content](#-main-content) - [🗒️ Notes](#️-notes) - [📚 References](#-references) - [🗣️ Paraphrase](#️-paraphrase) - [📰 Newsletter](#-newsletter) - [✅ Action Items](#-action-items) - [🧪 Hypotheses](#-hypotheses) - [📊 Methodology](#-methodology) - [❓ Questions](#-questions) - [🔮 Future Research](#-future-research) - [💡 Implications](#-implications) - [🧠 Learning Styles Assessment](#-learning-styles-assessment) - [📊 Accessibility Assessment](#-accessibility-assessment) - [🌍 DEIB Assessment](#-deib-assessment) - [✅ Logical Validity](#-logical-validity) - [🛡️ DARVO Analysis](#-darvo-analysis) - [🎯 Conclusion](#-conclusion) ### 🔗 References - Video: S-Tier MCP Servers for Developers — Syntax. https://www.youtube.com/watch?v=5az_4IwTwAE - MCP Registry (community index). https://github.com/modelcontextprotocol/registry - Sentry MCP Server. https://github.com/getsentry/mcp-server - Spotlight (local error tracking). https://spotlightjs.com/ - ElevenLabs Agents + MCP. https://elevenlabs.io/docs/developer-tools/agents - Context 7 docs loader. https://github.com/context-labs/context7 - Svelte MCP Server. https://github.com/sveltejs/experiments/tree/main/mcp - Chrome DevTools MCP Server. https://github.com/ChromeDevTools/mcp - Playwright accessibility testing. https://playwright.dev/docs/accessibility-testing - Cloudflare MCP tools. https://github.com/cloudflare (search “mcp”) - Stripe MCP Server. https://github.com/stripe/stripe-mcp - Mastra orchestration. https://docs.mastra.ai/ - OpenAI “Chat Toolkit” overview. https://platform.openai.com/docs ## 📝 Summary - The hosts rank practical MCP servers for daily development. - They show Sentry, Spotlight, Stripe, Chrome DevTools, and Svelte MCP. - They note docs ingestion with Context 7 and framework MCPs. - They stress reduced context switching and faster feedback loops. - Risks include context bloat and outdated model knowledge. - Bottom line: MCP servers compress setup toil and harden workflows. - Sources: video plus linked project docs above. ## 📖 Main Content - MCP servers expose domain tools to agents and editors. They reduce switching. - Sentry MCP manages issues, traces, and setup from chat or CLI. - Spotlight centralizes local errors. It assists debugging before production. - ElevenLabs shows voice agents calling humans, then feeding results back. - Context 7 fetches live docs. It curbs outdated code patterns. - Svelte MCP adds static analysis and an autofixer for Svelte 5 syntax. - Chrome DevTools MCP drives a headless browser for tests and screenshots. - Cloudflare and Stripe MCPs script common ops from language prompts. - Mastra composes custom pipelines and exports your own MCP servers. - Uncertainties: vendor coverage varies. Security and auth models differ. - Evidence: project READMEs and official repos provide capabilities. ### 🗒️ Notes #### ⏱️ [00:06](https://www.youtube.com/watch?v=5az_4IwTwAE&t=6) Framing - MCP servers now feel essential, not experimental. - Agents gain reliable tools for everyday coding. #### ⏱️ [04:12](https://www.youtube.com/watch?v=5az_4IwTwAE&t=252) Sentry MCP - Manage projects, issues, traces, and analysis via MCP. - Less context switching than using the dashboard. #### ⏱️ [07:10](https://www.youtube.com/watch?v=5az_4IwTwAE&t=430) SpotlightJS - Local error tracking with an MCP and a UI. - Useful for logs, traces, and early failures. #### ⏱️ [09:05](https://www.youtube.com/watch?v=5az_4IwTwAE&t=545) ElevenLabs agents - Voice agent called a developer, captured advice, and applied fixes. - Demonstrates tool chaining across modalities. #### ⏱️ [11:16](https://www.youtube.com/watch?v=5az_4IwTwAE&t=676) Docs ingestion - Context 7 fetches library docs to reduce hallucinations. - Beware context bloat and stale knowledge. #### ⏱️ [16:20](https://www.youtube.com/watch?v=5az_4IwTwAE&t=980) Svelte MCP autofixer - Static analysis flags syntax and suggests fixes. - Strong aid for new Svelte 5 patterns. #### ⏱️ [22:53](https://www.youtube.com/watch?v=5az_4IwTwAE&t=1373) Chrome DevTools MCP - Headless control for logs, screenshots, and performance. - Great for automated validation loops. #### ⏱️ [25:12](https://www.youtube.com/watch?v=5az_4IwTwAE&t=1512) Stripe MCP - Create products, prices, invoices, and coupons by chat. - Speeds internal support workflows. #### ⏱️ [27:56](https://www.youtube.com/watch?v=5az_4IwTwAE&t=1676) Mastra orchestration - Build pipelines; expose them as custom MCP servers. - Helpful UI for testing and branching. ## 📚 References - Sentry MCP Server. https://github.com/getsentry/mcp-server - SpotlightJS. https://spotlightjs.com/ - ElevenLabs Agents. https://elevenlabs.io/docs/developer-tools/agents - Context 7. https://github.com/context-labs/context7 - Svelte MCP Server. https://github.com/sveltejs/experiments/tree/main/mcp - Chrome DevTools MCP. https://github.com/ChromeDevTools/mcp - Cloudflare developer repos. https://github.com/cloudflare - Stripe MCP Server. https://github.com/stripe/stripe-mcp - Mastra. https://docs.mastra.ai/ ## 🗣️ Paraphrase - MCP servers connect agents to rich, purpose-built capabilities. Teams ship faster with fewer clicks. - Vendor MCPs and framework tools reduce setup time, raise quality, and standardize flows across stacks. ## 📰 Newsletter - MCP servers are quietly becoming table stakes for dev teams. - Try Sentry and Chrome DevTools MCP for triage and automated checks. - Add Context 7 or framework MCPs to keep docs current. - Build your own with Mastra when gaps appear. - Links: Sentry MCP, Svelte MCP, Stripe MCP, Chrome DevTools MCP. ## ✅ Action Items #### DevRel - Publish MCP quickstarts for Sentry, Stripe, and DevTools. - Record a demo on Context 7 powered issue fixing. - Track MCP adoption with example repos and dashboards. #### OSPO - Define MCP security guidelines and token scopes. - Approve vendor MCPs with risk reviews and logging. - Document standard MCP bundles per language stack. #### Community - Host office hours on MCP setups and pitfalls. - Share “before/after” time studies from migrations. - Maintain a curated MCP catalog with tags and owners. #### WordPress - Core: explore DevTools MCP for e2e checks in CI. - Playground: prototype MCP-driven test labs in-browser. - Training: add lessons on MCP workflows for contributors. - Meta: evaluate Stripe MCP for donations and events. #### The WPCC - Fund microgrants for MCP-based accessibility checks. - Sponsor guides that pair MCPs with contributor tooling. ### FAIR - Publish MCP policy templates for privacy and audit trails. - Add provenance notes to MCP pipelines and outputs. ## 🧪 Hypotheses - MCP-driven tests reduce regression triage time by 30%+. - Docs ingestion lowers library API errors by 40%+. - Stripe MCP cuts support task time by 50% for refunds. - DevTools MCP increases reproducible bug reports by 25%. ### 📊 Methodology - A/B teams with and without MCP bundles for six weeks. - Log time-to-fix, context switches, and error rates. - Survey perceived load and confidence weekly. - Audit commit quality with static analysis and reviews. ### ❓ Questions - Which MCPs deliver the highest ROI per seat? - How do we cap context size without losing fidelity? - What audit signals prove safe MCP usage at scale? ### 🔮 Future Research - Standard MCP telemetry for privacy-preserving analytics. - Playbooks for multi-agent orchestration with guardrails. - Benchmarks comparing MCP vs classic CLIs in CI. ### 💡 Implications - Teams should budget for MCP onboarding and governance. - Vendor MCPs shift effort from UI clicking to typed intents. - Custom MCPs encode institutional workflows for reuse. ### 🧠 Learning Styles Assessment - Reading level estimate: Flesch–Kincaid Grade 10.2. #### 👀 Visual learners - Present: diagrams of MCP data flows. - Improve - Summary: Scott and Wes break down the top-tier MCP servers developers are using right now. From browser automation to debugging superpowers, they explore how these servers are changing what’s possible in modern dev workflows 🔥 Be the ~14,700th person to join our super tasty newsletter https://bit.ly/syntax_snackpack Show Notes 00:00 Welcome to Syntax! 01:52 Brought to you by Sentry.io. 02:46 Submit your Spooky Stories! 03:37 Syntax San Francisco Meetup. 04:11 S-Tier MCP Servers. 04:17 Sentry MCP Server (not sponsored). 05:01 Pause, what the heck is an MCP server. 05:45 Back to the Sentry MCP. 07:07 Spotlight JS. 09:08 ElevenLabs MCP hack. 11:19 Docs Tools. 11:29 Context7. 15:55 GitHub MCP Registry. 16:10 Nuxt MCP Server. 16:25 Svelte MCP Server. 20:35 Cloudflare MCP Servers. 22:34 Chrome Dev Tools (Playwright) MCP Server. 25:17 Stripe MCP Server. 27:55 Mastra. All links available at https://syntax.fm/947 ------------------------------------------------------------------------------ Hit us up on Socials! Syntax: https://x.com/syntaxfm Scott: https://x.com/stolinski Wes: https://x.com/wesbos CJ: https://x.com/CodingGarden Randy: https://www.youtube.com/@randyrektor http://www.syntax.fm Brought to you by Sentry.io #webdevelopment #webdeveloper #javascript #syntax #syntaxfm #webdev - URL: https://www.youtube.com/watch?v=5az_4IwTwAE Welcome to Syntax. Today we're going tobe talking about S-tierMCP servers for developers. These aregoing to be MCP servers that work reallywell in general development flows tohelp you build stuff, to make things, tovibe code, any of that. And by no meansis this an exhaustive list or anythinglike that, but these are a handful ofMCP servers that I've used personallythat I have really enjoyed and that havemade my time working in code a lotbetter with agents. My name is ScottTolinsky. I'm a developer from Denver.With me as always is Wes Boss. What'sup, Wes?>> Hey, stoked to talk about this. Feelslike we've had a bazillion MCP episodes,but there's just like this never-endingwell of stuff to talk about, especiallynow like OpenAI just announced a wholebunch of their agents SDK. We'll have ashow on that, but it's never-ending. >> Yes, it's never-ending. And one thingthat I've personally felt is that likeyes, MCP to me always seemed like aninteresting idea, something that couldbe really advantageous to use, but wehave sort of reached a point where MCPServers are doing so much and so manygood things. They've gotten really goodthat I’m finding their utility to bemore and more something that I relyon every single day when I am usingagents to code. I actually think we shouldhave Kent Cods on the show. He's beendoing a ton with MCP servers and he andI chatted a whole bunch at the VSCode Insider Summit about MCP and histhoughts on MCP overall had reallyshaped how much I’ve dove intoadding these tools into my workflow. Lately, it has really paid off. Um, sobefore we get going, let’s talk aboutSentry. This show is presented by Sentrysentry.io/sintex.Sign up again for 2 months free. What isSentry? It’s the perfect place to fixyour bugs. Uh, make sure yourapplication is running smoothly. I’m goingto talk about a newer MCP server thatthey released recently beyond justthe already awesome Sentry MCP serverthat is an awesome S-tier developmenttool. So, that’s not even going tobe part of this sponsor read, right?That’s going to be part of the normalepisode because it’s that good.>> It’s just good. Yeah,It's just good. And I more and morerely on Century's tools to fix mystuff, man. Seir is their AI tool thatgoes through and finds the rootcause of your bugs. And I've had justsuch good luck using that tool. Socheck it out at century.io/sintax. Sign up for two months free. Tastytreat. All right, real quick before weget going. Two things. We need your webdevelopment horror stories, and we have ameetup in San Francisco on October27th. So, if you have a horror story forour yearly Halloween episode where wetell stories of people who dropped the database by accident,deployed a fix that DDoSed themselves,wrote a bad React hook, or accidentallyforgot the WHERE clause in a databasequery,oh man, these stories are awful. I wantto stick my head in my hands; I'm so stressedout reading these stories every year,but it’s so enjoyable. Please, pleasesend them to us. Go to syntaxfm/spooky,submit your story, or send me an email atwestboss.com.Send me a DM wherever. We would love toread it. We make everything anonymous.We’re not going to say who you are. But we’d love to read them off. And thenalso, on October 27 in San Francisco, we aredoing a meetup and we want to inviteyou. Uh, we are at Bear Bottle Brewing inSan Francisco. Go to syntax.fm/fm/meetup.You can grab some tickets. We’re goingto have some merch there. We’re going tohave some drinks there. We’re going tohave Scott there, myself, CJ, the wholeteam. It’s going to be a hoot, so comeon down. [Music]This is the end. End of the ad read. Let’stalk about S-tier MCP servers. Now, thefirst one that I have on here is theSentry MCP server. Um, the Sentry MCPserver is great because if youuse Sentry, which is again for trackingerrors and all that stuff, you want tobe able to integrate and use thistool to access your Sentry from your agent,from your CLI, in terms of like cloudcode, whatever you’re using. Um, allthese examples that I’m showing you onscreen today will be done with cloudcode, but if you cannot see, and you’relistening on audio, you’re not missing anything. Um, we’re mostly justkind of showing how they work. So, um, and we should say beforeYou even get into that, for anybody whohasn't heard, these MCPservers are a way to surface additionaltools toAI. And that could be coding, right, viacloud code or VS code or cursor. Itcould be via a chat app. Andit could be by simply making straight-upAPI calls to OpenAI orAnthropic, but you provided additionaltools to do things that it mightnot be able to do. It might not be ableto access specific things. Yeah, you're providing the AI withspecific tools to do specific things.So, I've loaded up the Sentry MCP andI've looked at the tools. There are 19tools available from find teams, findreleases, get issue details, get tracedetails. You can create a team throughthis. You can create a project, you canupdate a project. So what that means isthat oftentimes when you are in thepast, you're working with Sentry or anyof these types of services like this,what do you do? You go to their website,you create a new product project, youmaybe get an API key or a DSN inSentry's case, and then you runthrough the wizards or Sentry has aThe CLI does it for most platforms, andthe docs are great, but you're goingthrough and you're copy and pastingstuff from the docs. You're getting set up. You're having to click around. Withthis, you can create a project directlyfrom here. After that project is created,you can then search through issues. Youcan analyze issues with their SER AIplatform to help fix them. Ultimately,what that means is thatyou're spending less time leaving yourtools, such as cloud code or your editor,whatever you’re usingthis MCP server, VS Code, Cursor, orwhatever, you're spending lesstime leaving that to go to Sentry's UIto do things and more time just stayingwhere you are. As somebody who'sallergic to context switching,it's made a big impact for me. So that is the Sentry one because, again,we use those tools for me. Thisworks so well. But they also released anewer MCP server, which is theSpotlightJS, found atspotlightjs.com.Basically, this is like a localdevelopment tool where, like Sentry,I'm primarily using it mostly forproduction bugs and things like that. Where Spotlight is something that’scollecting your errors. The MCP serverin regards to this allows you againto get local errors; you’regetting local logs, traces, and eventsfor traces. And this is still brand new,but if you’re working in localdebugging, I know that I, personally,often have the MCP server or, uh,the agent running the processto read the console ormaybe even look into Chrome Dev ToolsMCP to read the console logs. WhereI’m copying and pasting it like a sucker,and there’s no additional context onhow often it happened or what’s the most commonerror. Um, yeah, this is awesome. Yeah. So, if you have this set up, youcould say, you know, read this error,whatever, and you’d be able to get thaterror directly from the MCP serverwithout having to,you know, instantiate other tools or,you know, run additional processes. So,I really like this. And on top ofthat, there’s also this SpotlightUI that you get. Um, so it’s not just anMCP server. There’s also this UI forbeing able to debug your app. This has become a>> like a local sentry.>> It's a local sentry for sure. And I'vereally liked this. Uh, andthey're working on it pretty hard becauseI gave them some feedback on theinstall process and sure enough, thislanding page has changed since I've donethat. So, uh, shout out to the Centuryteam for knocking this thing out of thepark. Um, can I talk about the 11 LabsSDK, or the 11 Labs MCP real quickly? This is not necessarily something thatyou'll do, but if you are the type ofperson who's like, "I don't really seethe point. I don't really seethe purpose in this type of thing." WhatI did is I took the 11 Labs MCP server,which 11 Labs does voice text tovoice, voice to text. Um, but they alsohave an agents SDK which is able tokind of spoof or make a phonecall, right? You can queue up an agentwith a whole knowledge base. Youcan give it a bunch of tools. You cangive the agents MCP servers as well,which is crazy. >> I guess that's like your MCP server canhave>> MCP servers inside of them. Totally. But the crazy thing was I had a bugin my code, and instead of asking, Ihad it set up in cursor. I said,instead of asking cursor to fix thebug, I said, "Call Wes and ask him how tofix the bug," and it literally rang myphone.>> Oh yes.>> He asked me what I would do in thissituation. I replied on the phone withwhat I would do. >> It transcribed it, put that back intocursor. Cursor then parsed the wholetranscript and then applied thefixes. Obviously, a stupid thing to do.But that just goes to show>> how, and I'm often just shocked at howyou don't have to write any code for any of these.They just all worktogether. Like so much of my career hasbeen wiring things together, and withthese MCP servers, it justmost of the time just language is wiredtogether.Yeah, exactly. So, I thought that wasa pretty funny and justif you haven't seen the benefit,try that. It's eye-opening asto how these things can worktogether. My cursor called me on thePhone. Cursor called me on the phone. Thatseems like a, uh,a nice little dev horse story. Iwas trying to fall asleep, and Cursorcalled me on the phone. All right. Sonext up is a class of tools that I’mcalling documentation-type tools whereyou are getting the documentationfor various libraries. Now, Context 7 isthe big one for this. And Context 7works super well because what you dowith Context 7 is you’re basicallysaying, “Give me the documentation forthis library.” Like, use Context 7 togive me the documentation. And what itdoes is it goes off and gets the IDfirst, and then it goes and gets thedocs. And that way, it has that in its context. So the agent is able to, inthe context of what it’s workingon, able to take in the documentation ofthe project that you’re working in. Nowthis can sometimes lead to context bloatwhere it’s bringing in a lot of too muchcontext, and I have some solutions forsome of that here. But I found that thisis one of the best things to dowhen you are finding that the agent orthe AI is writing code that is not inAdherence to the actual library. Um,because if you say use context 7 toget the up-to-date docs or whatever,it's going to be able to have that inits context and actually work on it. One thing that I've been doing lately isanytime I have a library, whetherthat's likeI wrote the um zero spelt bindings forzero sync, I threw that on context 7, andit's been so nice to be able to be likecontext 7 to make sure that you're usingthe documentation for this. It's so key,especially for libraries that the AI doesn'tknow about or it's just hallucinating it.Especially like sometimes it justhallucinates method names when itclearly has access to the types,which drives me nuts because like youshouldn't. Maybe you shouldn't even need docs ifthe types are good enough.Yeah, maybe. Yeah, sometimes it's justfor best practices or examples of how toconfigure it and whatnot. That that'sreally good as well. But like even justthe AI still writes old Tailwind allthe time and it's so frustrating. It writes old everything. Yeah. And youhave to explicitly tell it to usethese features, and that's a bit ofawe're kind of stuck in this 2023version of coding things unless youexplicitly tell it to use the newerversion of something. And like, I worryabout React — all these new Reactfeatures we talked about last week.Yeah. And like, it doesn’t know aboutspelt too. Yeah.There’s just so much of this stuff, youknow? For me, it’s like noteven that — it’s CSS, man. Like,the amount of things that I have tothrow into my context or I have tohave special CSS agents to write goodCSS, because AI models today — all ofthem. And people that I hate — whenthey leave comments, they just say AI,but they don’t even say which model. Well, that’s usually because it’s mostof the models, right? A lot of themhave the same issues. They all write dogCSS. They all add background colorto stuff that does not need abackground, right? This is you. Set the background of this input to bethe exact same as the background of thecontainer. Why? Yeah,they don't need that. That's just makingthis unmaintainable. Or you usedtransform translate instead oftranslate. Gosh, that's the MCP server Ineed! Do not ever use transh transformtranslate. Use translate already. Gosh. Yeah. What does Contact 7 have for likemodern CSS as well? Because we'vebeen blessed with so much new CSS in thelast couple of years, you know, and I don'tsee a lot of it having hugeuptake. And I think partially that'sbecause obviously of browser support. Ithink also partially just because, likethe AI just loves to absolutely positioneverything instead of using popover andanchor positioning. Let me throw an ass ton of JavaScript inhere when you could use the popover API.Like okay. Uh, I know I've been working onsomething for this Wes personally, Ibecause I’m just so fed up with the CSSthat it writes, and I have my own CSSsystems. You could obviouslytrain it on those systems, but I want, aslike a fundamental CSS fundamentals. Um, because obviously the models areall trained on bad CSS, because mostpeople write bad CSS. So it's a problem.But I personally want to solve this problem because it'sa problem. Now, in addition to context 7,there's something new kind of popping upwhere libraries are doing their own MCPservers. Um, I have the MCP registryfrom GitHub pulled up. Um, we'lllink this in the show notes.github.com/mcp. One thing that you'll see on here isthere is anMCP server specifically for KNX itself,working in Nux. And I haven't used theNux one, but it is by AntFu. So, I'mgoing to put it in S tier just based onthat alone. But the one that I have usedin this regard is the new Svelte MCP.This is their official MCP server,from the Svelte team. It's in thesvelte.dev. And all other libraries andprojects need to take note of howstinking good this thing is because onereally cool tool that it has besides thedocumentation is the bigproblem with Svelte is that they decidedto release a completely new syntax foreverything like the day before youKnow, models were released. So none of the models, like even modern models,anything that comes out, GPT-5, the new 4,they're all bad at spelling. Still, they'renot terrible; they're gettinga little bit better but they're stillbad at spelling. This fixes thatproblem because you can use SpellMCP to implement remote functions. It'sgoing to go off; it's going to get thedocumentation for remote functionspecifically, and it's going to use that towork on it. So it always has thatcontext—the same thing that you wouldhave with Context 7, but this one's verytargeted. It's going to know the exactfeature that you're asking it about. Butone of the coolest features in here isthe autofixer. The autofixer takes in a string of anentire component and then it returnswhat's wrong with this component. Thatway, the agent, the AI, whateveryou're working in, can then use thatinformation to update your projectlocally. So this is sending a string ofyour component, and it's saying, "Hey,you used onClick instead of onclick." It fixes the component thatrules. Need that.>> How does that work? Is it justlike providing a prompt to the AI,or is there literally a lintingserver hosted somewhere that it’slinting against?>> Oh man, that’s a great question for theSpelt team. I think it’sproviding a prompt. It’s sending yourstring of your component to aprompt somewhere. I don’t know. But umyeah,>> 'cause it also could be like, uh,running it through like the Doesn’tSpelt has a Spelt linter where itwill flag stuff that is maybe apossibly bad practice? >> Yeah. Yeah. Um>> along with how to fix it.>> Let’s see. Claude uses Spelt AutoixerMCP tolint.Um, let’s find a Spelt component. Let’sfind the — let’s do menu here and we’regoing to see what it does really quick.Basically, what it’s going to do is it’sgoing to, like I said, take a string ofthat. It’s going to send it off to theMCP server.>> The docs for this say it uses staticAnalysis to provide suggestions forcode. So >> cool. What it>> Thank you for looking it up instead ofjust guessing.>> Run all of that through a staticanalysis where it will look at all ofthe code, and there's likelythings in that as they loop throughthe tree of the spelt code that say, "Ah.">> I noticed that on click. Um, yeah, let'sthrow a warning for that as wellas a suggestion of how to fix it. >> Yeah, it returns an object withissues and suggestions. No issues.>> It probably just runs locally. So most of these MCP serversare simply just spinning up like npxcommands in the background. Some of themrun Docker containers, some of them arehosted externally. And>> this is a remote to run it. >> This is a remote one. Okay.>> Yeah.>> Cool. So I think other frameworks and libraries should take notethat rules. I know people don't havenecessarily the resources or whatever,but like that autofixer thing, man, I lovethat. I would love that for CSS. I love that for people who write Reactcode to do something that one line ofCSS would do. Like, hey, have youconsidered learning CSS beforeinstalling something? You know,>> Hey, maybe this is outdated orjust like a MCP server full ofdead simple examples that can then beprojected>> on your codebase, especially for a lotof the modern CSS stuff. Um, oneMCP server I love, or probably 15, is theCloudflare has>> like 15 different MCP servers for eachof their products. So they have one fordocs, which is good if you need docs on how workers run. But then you canalso use it just like an API tospin up a container or>> create a binding to a database in thisproject. So where you might havepreviously had to figure out whatthe wrangler command is that you needto type exactly to make this work, it’sso beautiful just to be able to type,you know, spin up a new D1 database andput it in my Wrangler bindings.>> Yeah, I love that because againotherwise, having to navigate theCloudflare UI does drive me nuts. Sometimes, I don't know about you, butas I use more and more of theirtools, I'm just like, man, I feel likethey could rethink some ofthis >> stuff. They're having the same problemthat Amazon had years ago, which waslike initially Cloudflare was deadnuts simple, and then as more productsget added,>> it's just like>> where do I even go for the... like Iwas looking for my tunnels the other dayand it took me like five minutesjust to find the link to see wheretunnels were. But their search wasactually really good, and I ended up justtyping in "tunnels" into the search. Um,so>> I happen to feel like it's a prank. They're like, "We have a tool calledtunnels, and we're going to make you gothrough>> countless tunnels and mazes to actuallyget to where you're trying to>> find where it is.">> I don't know that I could design abetter version of that. Like that'swhere>> noProper application designers come intoplay because they try to keep thingssimple while also surfacing all theknobs for when people want them, which is avery hard thing to design around. >> Yeah. Yeah, for sure. That is tough. Umthe next one here is the Chrome DevTools orPlaywright MCP servers. I’ve been usingChrome DevTools over Playwright MCPservers for doing the same type of tasknow since they were released. So, thisis a brand new one. If you haven’t heardabout it, Chrome DevTools have releasedan official MCP server. >> We did a show on it last Monday. So, youshould have heard it. If you’re listening, you should have heardit. This thing rules. Again, this oneis running with just an npx command, and itis firing this thing up. And let metell you, I’ve had such good luck withgetting into loops where I’m saying testthis in the Chrome DevTools MCPserver. That way, it can find consolelogs. It can find visual elementsthat are off with it. It’s takingscreenshots. It’s doing all kinds ofstuff, and you can see it work too. Soit’s not just happening blind inThe background. Yeah, like emulatenetwork, list console logs, click, drag,fill forms. I mean, there are 26 toolshere. Uh, getting performance. So, if youwant to evaluate performance, findinganimation issues or key framedrops, yeah,there are key frames, too. Uh, I don’t know if it’s accessible via the Devtools, but you know what is one movethat I really like in this? Ifyou’re working with an API that doesn’thave types and you don’t know what theAPI response looks like. Normally, whatyou do is just console log it andtake a look at itand then maybe you copy-paste thatobject out into some converter andconvert types for it. Um, but what youcan do in this case is simply just say,"Hey, run the fetch, console log it, seewhat you get, then write types for that."Then,change it to be like an error, make it intentionally break,and then look at what the response for that lookslike. Then all of a sudden you havethese beautiful types, and you can write a whole function that willYou. Yeah. And people who are allergic toAI stuff, like you got to be able to seethe utility in that. You know,you're doing that anyway.You're doing that anyway. Also, that'ssuch a bad waste of time. It'ssomething that you need, but you're likethe process of you going to dothat. Like that's not helping you. Solike, use the tool that just goes anddoes it for you. Verify, validate,all that stuff. But use the tool.It's great.This is the one that to mecouldn't be any more S tier: Stools—S tier Chrome developer tools MCP. Here's one that I used recently theother day that I was just blown awayby: the Stripe MCP server. My gosh, Itold Stripe to create a product,add it to my codebase or whatever, make abuy now button just for fun, and man, itwent through, and just instantly, it tookno time at all. I got everything right about it.I've used Stripe a lot myself, so it'snot like something that I'm justlike, "Oh, okay. We'll see if thisworks." No, I told this to create thisproduct in this specific way and within,You know, a minute ago, I had a workingproduct in my sandbox and I was just,wow, okay, this has saved me so muchtime again from going back and forth, hotpotatoing things from the UI to my code,adding IDs, whatever. The end resultcouldn’t have been better. I found theStripe MCP server to beshockingly super good. >> You know, another neat thing, like, ofcourse this is good for development,but I’m thinking about the amountof internal tooling I’vewritten against Stripe for refunds,partial refunds, taxes,payment, you know, being able to changesomeone’s address on an invoice, all ofthat. So much like subpar internaltooling as well, you know, like it’s notgreat UI. It works, you know,and my assistant knows how to useit. But I just look at this and thinkhow cool would it be to give her>> a chat box instead, or don’t even have aperson, you know, and like, of course,you just approve these things, but justas the request comes in, just queue up the chat and propose>> refund $6 in Canadian or something. Like that, and it just hit the button, andit will go off and do it for you. It'sjust that all that wiring up time is gonewith a lot of stuff.>> Time. Yeah. And, like I said,you can create products, createcustomers, list customers, createprices, invoices, finalize invoices. Youcan do so much stuff with this. Uh,create coupons, all this. The entireStripe dashboard is essentially here asan MCP server. I found it to be justtop tier, excellent, worked reallywell. If you're working with Stripe inany capacity, throw this in. Um, youwon't regret it. I found this to bereally, really great. The last one Ihave here is less of an MCP server andmore of a platform. Mastra, Mastra,however you say it, folks. This is atool that allows you to build orchestration pipelinesof potentially tools talking to other tools. But thecool thing about Mastra, and the coolthing about it is you can expose andcreate MCP servers with this. So, I canmake my own MCP server that maybe I'mwriting with just straight up JavaScriptto go and scrape pages or do this orthat. It scrapes a page. That tool then usesanother tool to save it to a SQLitedatabase. And then I have another toolthat’s exposed as an MCP server to useinformation from that database that I’mthen able to call for my application. SoI built a tool that was scrapingMDN and putting the things that I wantedto have from it in a SQLite database.Then I could use MCP from cloudcode to say, “Go grab this MDN doc.”It looks it up from my database, looksup the URL, then uses web search to loadthat page and reads some of thatinformation. So you can really gettotally customized with this thing. TheMCP. Again, the server that you’rewriting can just be straight upJavaScript. It can be the result of an agent. It can be the result ofother tool calls. And you can buildthese things all up and then expose itas an MCP server. This isn’t like abrand new tool; it’s new-ish and it’sreally super good. Another thing thatsome people might like about this isthat there is UI in all of this stuff.So as you’re building it, when you’retesting these things, you can clickaround, you can see the results, you canSee the returns? You can runthings in parallel branch loops. I mean,it's a very powerful tool even if you'renot using its most powerful features. This thing rules, and I've beenhaving a lot of fun building my ownstuff with it. Yeah, OpenAI actually just releasedsomething yesterday at the recordingthat's very similar to this as well,which is just wiring things up. Youknow, you can click and drag thingstogether. I still think, though, that if this andthat kind of dropped the ball becausethis was their game, you know, butyeah, this is really cool justto be able to drag and drop and click things together as you need it. I was looking into this as wellbecause a lot of people are thinkinglike, "Okay, I need something like thisfor my business," and OpenAI is trying toget you to bring that into ChatGPT or you can host your own. They released this thing called Chat Kit,which kind of seems similar to this, butyeah, a lot of people want to build thesecustom flows for themselves, but alsogive them topossibly the non-technical audience that. They can just wire things up.Yeah, this to me feels like firmly a dev tool,but then you can expose it as an MCP serverto then connect to whatever application you're using it in.So there's so much here. Um, this thing isreally pretty neat, and I found workingwith it to be generally, um, you know,there are some iffy things here and there in the docs.It's some new stuff here and there, but like in general,I found this to be especially for developers to feel like,wow, this is a really cool tool.I’ve been really, I don't know how much you've had to dealwith this, but I've been really getting into context management.Um, people call it context engineering sometimes. And it's like, how do you give the AI enough context to do its job,but not too much? Because the more and more you pollute that context,the more and more you get bad results, hallucinated results.Uh, you fill up the context and then it has to summarize.And then, man, I was doing so much with agents and sub-agents thatwere first calling and reading large MD documents,and then it just like wouldn't be able to do anything because thatThe context window would fill up so quickly. So yeah, a lot of interesting stuffhere. These can solve some of thoseproblems. It's probably worth an episode ofits own. Uh folks, is there an MCPserver that you use in your developmentthat really, really helps youout? Is there something that we didn’ttalk about that you feel isnecessary in day-to-day development?I want to hear from you because everysingle time I find a new one of these,like the Stripe one, that makes mereally excited. I just think, "Man,this is a really cool placeto be where I’m no longer having to golearn and explore some UI. I’m nowgetting to interact with it in thelanguage. I need a product. It needs tolook like this. Give it to me now. Youknow, love that. Beautiful. Thanks so much for tuning in. Catch youlater. Peace., S-Tier MCP Servers for Developers, Syntax, 01k818jjqyac16hemb3z52q4hm.