10 Cities. 4 Countries. One Unexpected MCP Lesson.

Demetrios: [00:00:00] All right, so at the beginning of April, I went to New York for the MCP Dev Summit, and while I was there, I got the chance to record a few podcasts with attendees that were at the event. This is one of those conversations. Hope you enjoy.

Ben Morss: I think the main thing is that people don't quite understand what it is that easily.

Ben Morss: People think it's some way of mixing AI into your API in some way, like your API's gonna change what it does, and magically it's gonna sort of do new things, which of course isn't what it is. It's AI using APIs. There's a lot of questions you get that I can't answer that well because they're about other ways to actually, like inject prompts back into the LLM or, you know, other things.

Ben Morss: But I think part of it is just that if you're, uh, using an MCP to use your API, it's good to not expose too much of it, and it's good to also have your API be very secure in the first place.

Alex Salkever: Hi, I'm Alex Salkever, and I'm here at, uh, MCP Dev Summit [00:01:00] North America in New York City. I'm here with Ben Morse of DeepL. Hi. Um, which is a... Tell us what it is, please.

Ben Morss: What is DeepL? DeepL is a, a AI, uh, language translation company based in Germany. We've got about 1,000 employees, so it's like a popular, like, phone app and web app and such thing in countries like Germany and other European countries, and also Japan and even Korea.

Ben Morss: Not as well known in North America, so they hire people like me to be here in North America and say, "Hey, we exist. Hello. Hi. Here we are." Good

Alex Salkever: plug. Love it, love it, love it. Thank you. So

Ben Morss: the

Alex Salkever: reason I wanted to talk to you though is your- Yes ... recent MCP roadshow. 10 countries in 10 days, I believe it was?

Ben Morss: It wasn't quite 10 countries, but there

Alex Salkever: were- Okay, well, yeah, tell- Or

Ben Morss: 10 days, but

Alex Salkever: there were- How did this, how did this come about, this mad plan to spread MCP to the world?

Ben Morss: Yeah, it was kind of, it just kind of grew organically. So when I got to DeepL and became the developer advocate person, I had this remit to make it so people knew what DeepL was in North America, and so developers knew [00:02:00] we existed. So I had to get out there in communities and just find people, and so I was looking for ways to do that.

Ben Morss: I thought, "Oh, here's a hackathon in San Francisco." Like, this was like May 2nd to May 3rd. I was gonna say, 10 cities and...

Alex Salkever: Sorry,

Ben Morss: not countries. 10 cities, yeah. Uh- It was 10 planets and 10 light years. No, it was, uh, it was 10 cities, I think. Very cool. 10 cities, 10 talks, and four countries, I think. Uh, yeah, we're gonna go into the US, of course, Canada- Yeah

Ben Morss: France, Germany, I think those are the countries. Uh, there's demand in more places too, certainly for this. So, um, anyway, so I thought, "Okay, we'll be bold. We'll go to San Francisco and sponsor a hackathon there." And I was, uh, you know, I was following AI and following agentic AI, listening to lots of podcasts and trying things out.

Ben Morss: But I didn't know what MCP was at all. I just kept hearing about it. I thought, "What are we gonna do to go to San Francisco with all these cool, like, AI hipsters, you know, with all their cool AI companies based in the Bay Area?"

Alex Salkever: I was like, "Oh, it's Tuesday. There's 75 hackathons," so

Ben Morss: yes. Exactly. Yeah, and like, I know this deep at all what are we gonna do.

Ben Morss: But, you know, I'm pretty brave sometimes, uh, so I figured we would just go there. And then, [00:03:00] um, I heard about MCP and I thought, "We should build an MCP server. This is, like, the lat- latest hotness. People are using this stuff." Just on the fly,

Alex Salkever: like YOLO MP- MCP server.

Ben Morss: That was the idea, and then there was an engineer who knew about it too and wanted to try it out, and, uh, his manager thought it was a good idea, so we just built an MCP server.

Ben Morss: Mostly him, his name is Akash, and he did most of the work. And, uh, we just kind of helped him, like, debug a little bit, and then it seemed to work. So you did this

Alex Salkever: in, like, essentially, like, a week or something, right? I mean- It was

Ben Morss: pretty quick, yeah. Okay, cool. Which you can do with MCP, as we've learned. It isn't that hard to actually use if you know-

Alex Salkever: Right

Alex Salkever: how to use it.

Ben Morss: Uh, which we didn't, of course. So there were some confusing things that happened, some confusing bugs. But basically, we finished it. I went down to this hackathon, gave my talk at the beginning like, "Hey, you got an MCP server?" And everyone just figured one of the many AI companies that existed.

Ben Morss: People accepted us as one of their own, and it was actually pretty successful. People didn't use it much in the hackathon, honestly. They kind of used the API instead. But having, having it meant that we were, like, a sort of a player in the scene. It was kind of like wearing the right shirt or the right glasses.

Ben Morss: We showed we knew what was going on with MCP. That's how it all began, just kind of as a spontaneous idea [00:04:00] to fit into a hackathon.

Alex Salkever: And so how did it progress? Because it sounded like you learned a lot From leading essentially 10 MCP hackathons in four countries over a very short span of time. Yeah. I mean, I'm kind of curious, A, what you learned, B, what you saw that was different in the audiences, the-- or even the tenor of how people were approaching MCP, how it was different.

Alex Salkever: You know, like what, what did you learn?

Ben Morss: I just thought it was a really good topic, and I just got really interested in the topic myself. I mean, tech these days is really fascinating. It's changing all the time, and AI is so interesting, and I got really deep into actually how it worked. I thought, how could this possibly happen?

Ben Morss: Like this is an LLM, it just like says stuff. It talks to us. We know really what it is. It takes in tokens as input, puts out tokens as output. How does it like use tools? And none of the explanations I saw talked about how it actually worked. They were all kind of similar and talked about the USB-C for AI and didn't talk about- Right

Ben Morss: what actually happened. So I did a lot of thinking about it, investigating, asking friends who are more deeply involved in AI than I was, and so I really understood it well. And then I thought, well, I have to give a talk in June in New York. I'll give a workshop at MCP 'cause I [00:05:00] think it should be a thing everyone understands how to do, so this would be a good reason for me to figure out like in a couple of weeks how it really works well enough to actually teach people how to, how to use it, which is a big challenge to actually teach a thing you don't really know at all.

Ben Morss: But again, you try to go for challenges in life, and this makes us learn things. So I realized it wasn't that hard to actually make a server. Like you could actually make a server in seven lines of Python. You could explain it actually pretty easily. I found like the inspirations for it, like the ToolFormer paper from a couple of years ago that talked about how to actually, um, just, uh, training the model with, uh, data that wasn't just data, but also these faux function calls that then could be spat out by the model later on as actual function calls intercepted and then used as actual function calls.

Ben Morss: So, uh, I gave this workshop, and it was surprisingly successful. Like the room was packed. People had a lot of questions. They actually listened. I think for the most part, they didn't ignore me as much as they might normally ignore me. Uh, and afterwards people were like, the next day, "I built an MCP server.

Ben Morss: I did it. Here it is." So I realized there was demand. And you're kind of like, "This is

Alex Salkever: great." Then you're like, "Oh."

Ben Morss: Yeah. [00:06:00] People really were interested in the topic, and it's a nice entry point to learning about how to use AI. Yeah,

Alex Salkever: yeah.

Ben Morss: Because, uh, I realized you could build a server pretty quickly. I made a sample server that was just basically, it told jokes.

Ben Morss: So, uh, the, the simplest tool just returned the same joke every time. So very easy to understand, you know, just return the same string every time. And the jokes API I found was very easy to use. An API key was necessary. You can just call the API as a REST API. So I made more tools to get various kinds of jokes, and I thought it was pretty easy to understand how it worked.

Ben Morss: It didn't look like complicated code or anything. Once you install a fastMCP or your favorite client library, you could just kind of do it. So I made some sample servers like this, and I started using these in more and more workshops because people always wanted to hear about MCP It was a good topic for talks.

Alex Salkever: Obviously. Yeah. So like over the next seven, eight, I mean, like what changed? What progressed? Was, was there like-

Ben Morss: Yeah ...

Alex Salkever: sort of a, you know, uh, things that at, at the end that were very different or? I

Ben Morss: did a bunch of Google Dev Fests actually, because I used to work at Google. I had friends there, and so they had me come in October.

Ben Morss: Actually, my, my, my, uh, [00:07:00] my indie rock band played opening the show. And then I gave an MCP talk after that, which is kind of a funny combination possibly. Um, what changed? Um, there were... Mostly it was just there were cases that people were more interested or less interested. People needed to know about it more in some cases, less in some cases.

Ben Morss: But I think the demand is still as high as it was before. I did my last workshop on pure MCP in February in Philadelphia, and, uh, it was for an audience of AI specialists who really understood how to train models more than I do. And they were still interested in the talk and didn't know these things about how it worked, and I mean, I think the main thing is that people don't quite understand what it is that easily.

Ben Morss: People think it's some way of mixing AI into your API in some way. Like your API's gonna change what it does, and magically it's gonna sort of do new things, which of course isn't what it is. It's AI using APIs. Um, so I think they like the idea that you can actually get the AI to act in a rational way.

Ben Morss: Like use one of these tools over here, but use the API in other kinds of ways. It's restricted, you know, it's, uh, it's, [00:08:00] uh, discreet. It's not as sto- stochastic. Um, but to your question, what actually has changed since then? Not much actually has changed, to be honest.

Alex Salkever: That's cool. So essentially the cert- the, like the protocol people are starting to get it, and the, the questions are better and-

Ben Morss: The desire to learn is still as big as it was before.

Ben Morss: The questions are kinda the same questions. Like-

Alex Salkever: So what are like the top five questions that people get, or the misconceptions or the- Yeah ... I mean, if you were gonna redesign the course from the ground up, what, what would you do differently?

Ben Morss: I would like show how it worked by actually showing an example.

Ben Morss: Actually, it's not that hard to do. You actually can go to Claude or your LLM and just tell it, "Okay, now you can use tools with this protocol over here," and put in some fake XML or JSON, and it will say, "Okay, I'll do that." And it'll begin spitting out things using those tools. You could then say, "What's five times eight?

Ben Morss: Use your new math tool." It uses this math tool it already knows about, and in- instead of spitting out 40, it says, you know, like, uh, angle brackets math five times eight, close angle brackets math. So now the LLMs are good at actually using tools and smart about it. I think I try to just talk about how it actually works, [00:09:00] you know.

Ben Morss: It actually gen- generates a prompt, and the prompt is a thing which gets in the system prompt, and this explains how all the tools work. This is why your context windows can get eaten up, because you have a lot of stuff in there. I would explain how it worked and what it is. People wouldn't be as, uh, surprised by what it actually is.

Ben Morss: Uh, so I still get questions about that, like how it actually works, and questions that reflect What really is going on. Um, but more common questions I get are about security.

Alex Salkever: I was gonna say.

Ben Morss: So

Alex Salkever: what do you tell them?

Ben Morss: I tell them be careful, 'cause it's kind of like shareware. I mean, if you're downloading someone else's code to your laptop and letting it run, then be careful about that kind of thing.

Ben Morss: Sure. But

Alex Salkever: on the design side- Yeah ... what are the con- 'cause I mean, uh, you know, the project has a long security best practices doc. Mm-hmm. What do you tend to start with in, in helping people?

Ben Morss: There's a lot of questions you get that I can't answer that well because they're about are there ways to actually, like, inject prompts back into the LLM or, you know, other things.

Ben Morss: But I think part of it is just that if you're, uh, using an MCP to use your API, it's good to not expose too much of it, and it's good to also have your API be very secure in the first [00:10:00] place. In the end, it comes down to API security. You can use the same kinds of things, I think, to secure your API and secure your API against LLMs using your API.

Ben Morss: And then, yeah, look for those best practices and authorization. Um, make sure the people can't just, like, uh, easily spin up gigantic queries that will, you know, slow down your, your servers and things like that. Honestly, I don't have the best advice for security. I think it's still a thing people have to deal with themselves and look into themselves.

Ben Morss: My best advice is still just, um, make sure your stuff is secure. You know, be careful about it, and make sure you do the usual things, sanitizing input, all those kinds of things you usually would do. Look out for authorization problems, those kinds of things.

Alex Salkever: And so as you kind of traveled from- Yeah ... one of these to the next, um, did you see differences in the way people thought about it?

Alex Salkever: You know, like the AI experts versus the noobs or people that are sort of full stack guys or, you know, you- Mm-hmm. Uh, what were the differences in the questions you got from the different audiences?

Ben Morss: I couldn't always tell, honestly. That's a good question. I should be able to tell, but [00:11:00] it's just different places had different kinds of questions and different kinds of levels of interest.

Ben Morss: I gave one talk actually at a, it was actually at a business school in- Oh, really? ... Canada. It was in their Google Dev Fest, but the students weren't that interested. The professionals were, were more interested. Um, best talk I gave, I think, at all was, uh, at API Days in Paris. The room was really full. People were outside the room, and they were really just curious about what this thing was and how they could use it themselves.

Ben Morss: I think we overestimate how people actually might know things, because people that are really involved in a technology really think everyone knows what they're doing. But everyone else out there is just kind of like, "Oh, I saw this thing. I saw an article about this thing somewhere. But what is this thing?

Ben Morss: I don't quite know what it is." Unless you actually use it, you don't really get what it is. So if you were going to try to summarize like, uh,

Alex Salkever: you

Ben Morss: know-

Alex Saltkever: Yeah ... the, the, the quick start guide in, I don't know, five minutes or whatever, what, what, what are the sort of the principles you'd go over?

Ben Morss: I would just show how to build a really simple MCP server.

Ben Morss: And yeah, talk about the, the principles first of like basically what it is, it's a, a way you can get, [00:12:00] uh, an LLM to actually make these faux function calls, and you then control the function calls yourself. It calls your server, your server's working over here. It's just basically functions. It can call your API or not, but that's all it is, is basically you write functions that then the MCP server can go ahead and call if it wants to.

Ben Morss: And then, uh, show how to build some simple ones, you know. Simple, simple things. Once you get the simple ideas down, the rest can follow from there. What I found harder, though, MCP apps and other kinds of new things, they don't always work for me quite yet, and they're harder to explain. Like the protocol for MCP apps is more complicated, MCP UI, this sort of thing.

Alex Salkever: Yeah.

Ben Morss: I have not found a simple way to actually explain how that works or to even show how that works.

Alex Salkever: So were you involved with the design and construction at DeepL of the MCP server? I mean, did you kind of watch the process?

Ben Morss: Yeah, I kind of maintain it myself now.

Alex Salkever: So, so- So what, what were the conventions around that?

Alex Salkever: 'Cause I mean- Hmm ... it's kind of interesting, I, I mean, the idea of a language server over MCP. Yeah. Uh, you know, like w- sort of why or what's different or, I mean, and how do you design it, uh, you know, so that it's [00:13:00] different than a straight up API or something like that?

Ben Morss: Yeah, it's not that different. I mean, it's mostly about trying to find the most commonly used functions of the API and make them really easy to use.

Ben Morss: Like, uh, s- send translation texts, you know, sending texts, getting texts back And just, uh, trying to make sure you can look for errors, like are language codes used correctly? Do you require a country code over here or not? And give helpful information to the LLM so it knows what to actually do. And when I first did this, I remember it would have failed because it would use language codes incorrectly.

Ben Morss: It would use lowercase instead of uppercase sometimes and things like that. So it- So it's

Alex Salkever: not that forgiving yet.

Ben Morss: But now it's much more forgiving because, one, we made it more forgiving by giving more information in the annotations- Yeah ... type annotations and so on. And two, LLMs have just gotten smarter about what they try.

Alex Salkever: So when you started, what were, what were you working on? I mean, what was the LLM in the back, what you were piping into?

Ben Morss: I usually use Claude because Claude Desktop was the most fun thing to demo in. Okay. And I started to make, make conversation with it. It would try things out and try to be helpful. But it's gotten much better at this stuff.

Ben Morss: Like last talk I gave on this, I, I did the talk at DrupalCon last week, and I wrote an MCP [00:14:00] server that, um, would, uh, look for translations existing in your Drupal site. And if there weren't any, it would suggest making new ones for you and send those over for you. And it began just doing all this stuff by itself.

Ben Morss: I gave it a couple of tools, and it would, it would use DeepL's MCP server, which already is in my Claude Desktop, and use my one I made for Drupal, and it would suggest all kinds of things. "Do you want me to translate this into Danish for you? And here it is." And it also would comment on translations because it also knows translations itself, being, you know, also Claude behind the scenes, so it can comment on the translations in other kinds of ways.

Alex Salkever: Yeah. Did you-- I mean, so essentially you had it almost like co-browsing and like- ... "Hey, look at this Drupal site," and- Yeah ... you know, and do, sort of pore over the CMS and look at, uh, sort of all the meta text or whatever and, and tell me what you think it... I mean, and then it went on its own. It's like, "Oh, by the way."

Ben Morss: Yeah. I gave access to look at the nodes and see what was in the nodes, the article nodes, other kinds of nodes, and it would just look at these things and start making suggestions. Which I didn't expect at all. I thought it would require a lot more human intervention, but it really tries to be helpful.

Alex Salkever: And, and that's, like, something that in the past, weren't there specialized [00:15:00] tools that you would u- you would, you know, use for translation versus, you know, sort of generic, uh, you know, LLMs? Mm-hmm. I mean,

Ben Morss: I'm- That was the idea exactly, because people use these things called CAT tools, where they're these large things with the often very old looking interfaces that are kind of like a, a scarier looking version of Excel, where they send off stuff for machine translation as a first draft.

Ben Morss: Then the translator will then modify them as they want to, and it can go and find whether text you've translated before is present in new text, and it can try to suggest more things for you. So I thought if you're actually a translator working for a Drupal site, you're not gonna wanna go and probably walk in and, like, start typing things into Drupal because it can be a little confusing.

Ben Morss: Maybe you're gonna wanna use your favorite LLM and just talk to it. You know, use Claude- Yeah ... or ChatGPT or something. And it was surprisingly effective at that Like it was really pretty friendly

Alex Salkever: So you almost discovered a new product category for DeepL without realizing what you were doing.

Ben Morss: Yeah, maybe we'll actually make this into a product people could use.

Ben Morss: I mean, I think in these MCP apps you could actually... What it can't do now, the server I made like in a couple of days, is you, uh, you can't really modify things yourself [00:16:00] that easily. You can, but it's not that- So it's mostly read-only

Alex Salkever: and-

Ben Morss: It, it's like, you know, it'll say, "Okay, here's the translation for you.

Ben Morss: Do you wanna use this or not?" You can type a new translation or you can copy and paste it and change things. Should be more like an interface, more like a CAT tool where you have little boxes and, "Here's the translations. Do you like these things or not? Feel free to modify these things. Send them all off as a batch to the Drupal site."

Ben Morss: That'd be a better way to do it.

Alex Salkever: And, and from my understanding too, or the audience's understanding, um How is DeepL's translation layer differentiated or s- or, or, you know, better than what you might get out of the frontier LLMs right now?

Ben Morss: Yeah, th- that's an interesting question because DeepL first came out in 2017.

Ben Morss: They had the best translation anywhere because they had their own models, they trained to do these things with linguists' help, and it was better than Google Translate, better than anything else. LLMs came out with pretty good translation just kind of out of the box.

Alex Salkever: Yeah.

Ben Morss: So DeepL has now trying to stay ahead by just taking models, building its own models, fine-tuning models, and just doing more and more testing with linguists, more and more fine-tuning.

Ben Morss: And, uh, studies and [00:17:00] tests do show that DeepL is better for most language pairs in most cases, often substantially better, but it's more of a race now because LLMs can do pretty good translation.

Alex Salkever: Yeah.

Ben Morss: If you want very good translation, you should use, uh, like a company like DeepL that does it, you know, only this kind of stuff.

Ben Morss: People often say they prefer it and people will use this because it's just better. You're translating stuff kind of casually for fun or just a few strings on your site, you could probably use an LLM out there and do it that way.

Alex Salkever: So if anybody wants to use the DeepL MCP server- Yes ... they can basically integrate it into, like, as a front layer in front of anything that they're doing, um-

Ben Morss: Yeah, just stick it in there, try it out, and your LLM will know to use DeepL usually instead of its own translation these days because it will know there's a tool for that.

Ben Morss: They're kind of biased towards using tools as opposed to not using tools.

Alex Salkever: Yeah.

Ben Morss: And it just gives you a higher quality of translation. If you have things where your translation quality really matters to you, you know. It's pretty cheap actually to DeepL.

Alex Salkever: So are people using-- Were people actually building things, MCP apps or whatever, with the DeepL MCP during this tour?

Alex Saltkever: I mean, did you start to [00:18:00] see people try to do translation stuff and-

Ben Morss: I can't tell if they're coordinated or not. We've seen growth quite a bit in our DeepL MCP server usage.

Alex Salkever: Yeah.

Ben Morss: It's going, growing very-- It's still very small compared to the actual API, which is very large of course, but it's growing very fast.

Ben Morss: I hear from folks sometimes who are building things with it and they're making various things with it, and I think they just find it more convenient than using the API sometimes because our tools are more carefully defined, you know. Like, here's this translation text, here's translating a document, and there are more, more guardrails around these things, so th- I think it's just easier to use.

Alex Salkever: So, like, essentially the flexibility of MCP was something that as you were traveling around, people were-- It's just much easier for, for people that don't wanna have to think in terms of the structures of an API or REST API-

Ben Morss: And sometimes it's the reverse because an API may be a whole new thing. Like, what is this API?

Ben Morss: How do I use this thing? MCP, I think it's just sort of simpler. Here's the tool. You know, you give this tool these two or three parameters, and then it gives you the results you're looking for.

Alex Salkever: Very cool. Um, so when's the next tour?

Ben Morss: That's a good question. The last talk was just last week, [00:19:00] but I'm actually trying to focus on making some scaled content, like writing more of the traditional DevRel things, like more documentation, make some videos, things like that.

Ben Morss: So I'm taking a little break from traveling out to do those things. And, uh, when I go back out there and talk again, maybe more about agentic AI and more of those sorts of topics that are also- Yeah ... of course, hot topics. You know, building agents, which is super hot, as you know. But MCP, I mean, people still don't know what it is in most cases.

Ben Morss: Outside of New York, San Francisco. Or people that are just following these things and who are working with it and have tried it themselves. Yeah. Those people are still like, "What is this thing?" So I think the demand- Outside the AI

Alex Salkever: bubble ...

Ben Morss: we always forget the demand is still there for just, you know, introductory content.

Ben Morss: It's always there for everything.

Alex Salkever: So the next time you do a tour, what, what, what are you gonna do differently when you design your MCP curriculum?

Ben Morss: I'm gonna try to make an MCP app this time and see if I can go all the way to making apps and using the new fancier features and seeing how that goes. It's a good question.

Ben Morss: I feel like what I had did before, um, how can I improve it? I feel like actually if I could get people to do more coding themselves, that'd be nice because then they learn [00:20:00] things for real. I would have workshops, people often would come in without a computer and want to just listen to a talk.

Alex Salkever: Right. So they didn't do hands-on as much.

Ben Morss: Actually, when I first did it, I thought I'd do a workshop, and people just came and they were looking at me like, "Okay, talk to us." So I made it into a talk spontaneously, and they did it later in some cases.

Alex Salkever: Yeah. It's actually... Can actually

Ben Morss: try it themselves, I think it's probably better. So I think if I was gonna do it again, if I do do it again, I'll try to get folks to actually really write some code.

Ben Morss: I know that you can use Python, you can use TypeScript, you can use your favorite language. You can do it pretty quickly, as long as you've got something installed on your computer, which is an LLM, which is an AI client, which maybe people, people don't, haven't got yet. If they have an AI client, they can do it.

Ben Morss: You can install your library. You can just go and do it, and it feels, it feels like magic actually when LLMs are actually, like, using your tool, and you see it doing that and then responding with your information you've given it. It's really kind of cool. I,

Alex Salkever: I mean, I, I personally, I'm not a coder. I've been...

Alex Salkever: Well, I'm, I'm a dangerous coder, right? Like, I could code stuff you would not- Everyone, everyone's

Ben Morss: a

Alex Salkever: coder now, so ... you would not want me to put it in prod. But [00:21:00] yeah, it's just blown me away how, you know, you can essentially go into Claude Code, give, give it a spec-

Ben Morss: Yeah ...

Alex Salkever: it'll, it'll recommend a tool. Or if there's MCP servers there that'll make sense, it'll recommend them.

Alex Salkever: You know, it's a, it's, it's, it's really, really different.

Ben Morss: Yeah, watching, like, again, like Claude or ChatGPT or something actually use your tools and then call two or three of your tools, like I could do a thing. It's, when I give a demo, which is like, "Okay, now I'm gonna have it use a translation tool. Now I'm gonna ask it to translate this thing, uh, but, like, find a joke, and then translate the joke."

Ben Morss: Well, call two tools, and people are on the br- edge of their seats. Now we're gonna have it actually, like, tell, tell a joke, translate it into French, and put it into a spreadsheet. Three tools, and you watch as Claude spins, and then actually usually it does it right. So it's just, it's exciting. It's oddly suspenseful and gratifying because- Super awesome

Ben Morss: it usually works.

Alex Salkever: Thank you very much for coming on the podcast. Thank you also. Uh, tell the audience one last time, you know, like who you are and what you do again in case they're catching a tail end.

Ben Morss: In case you forgot, uh, or missed the beginning 'cause you, what- whatever reason, you were doing something else, uh, I am Ben Morris.

Ben Morss: I'm from DeepL, a developer advocate, and we're [00:22:00] talking about MCP servers, how they work, and DeepL is a language translation company that uses AI and makes its own models to make language translation as good as possible.

Alex Salkever: Thanks for coming.

Ben Morss: Thanks for having

Alex Salkever: me. Enjoy MCP Dev Summit.

Ben Morss: I will. Thanks.