WEBVTT NOTE Sentence-level transcript of https://www.youtube.com/watch?v=sAOBXCDiDOs NOTE One cue per sentence. Cue ids are the line anchors on /transcripts/sAOBXCDiDOs.html. A cue ends where the next begins, or 2 s after its last word. s1 00:00:01.680 --> 00:00:02.560 Hello y'all. s2 00:00:02.560 --> 00:00:03.560 My name is Pedro. s3 00:00:03.560 --> 00:00:05.960 I'm the co-founder of Manufact. s4 00:00:05.960 --> 00:00:09.600 And uh today I want to talk to you about MCP apps. s5 00:00:09.600 --> 00:00:11.760 Specifically, we're going to talk about the primitives. s6 00:00:11.760 --> 00:00:15.600 So, how these apps are built, how they work, and what they allow you. s7 00:00:15.600 --> 00:00:17.560 How to distribute MCP apps. s8 00:00:17.560 --> 00:00:23.440 So, what is behind the discovery mechanisms of MCP servers and apps in general. s9 00:00:23.440 --> 00:00:30.520 And uh why I think you should care about this because this is how all software will be used. s10 00:00:31.360 --> 00:00:36.480 So, I'm sure most people here listening to this talk know what an MCP is. s11 00:00:36.480 --> 00:00:41.160 MCP has been around for quite some time since 2024. s12 00:00:41.440 --> 00:00:51.760 For a full year, it was a it became very uh say frequent talk amongst developers and companies that were rushing to build these MCP servers. s13 00:00:51.760 --> 00:00:53.760 MCP apps are less familiar concept. s14 00:00:53.760 --> 00:01:03.360 They've been around also for some quite some time uh more or less uh since the uh the end of 2025. s15 00:01:03.360 --> 00:01:10.720 But when I talk to companies and people in general, I see that many people don't understand what they are and don't understand they can build them. s16 00:01:10.720 --> 00:01:15.600 And uh specifically, they don't know how to distribute them. s17 00:01:15.640 --> 00:01:21.520 MCP apps, but also they don't know the new way to distribute MCP servers as well. s18 00:01:21.520 --> 00:01:28.520 So, I hope by the end of this talk you going to know about this and you're going to be ready to build your first or iterate on your MCP s19 00:01:28.520 --> 00:01:37.040 uh server and app and you're going to be able to share it with the world in a more efficient way brings you more customers and more users. s20 00:01:38.360 --> 00:01:42.240 A little words about me and the Manufact uh my company. s21 00:01:42.240 --> 00:01:49.840 Uh we build open source SDKs and the tools for MCP and MCP cloud uh of the Manufact. s22 00:01:50.680 --> 00:01:57.901 Open source SDKs uh by the name of MCPUs uh allow developers to build servers um s23 00:01:57.901 --> 00:01:58.280 [clears throat] s24 00:01:58.280 --> 00:02:00.600 clients and agents in an easier way. s25 00:02:00.600 --> 00:02:08.920 So, we provide an abstraction over the official SDKs that allows developers to ship faster without worrying about how the spec works beneath. s26 00:02:08.920 --> 00:02:15.320 We have 8 million plus downloads across our SDKs, and we have 10K stars on GitHub. s27 00:02:15.320 --> 00:02:20.440 Also open source, uh the other product is the inspector. s28 00:02:20.440 --> 00:02:32.560 Uh it's an open source inspector, again, uh something comparable to the official inspector from the model context protocol maintainers that allows you to test these MCP servers and apps specifically s29 00:02:32.560 --> 00:02:34.440 uh on your local machine. s30 00:02:34.440 --> 00:02:39.120 Once you build, once you test it, we have built the cloud for you to ship. s31 00:02:39.120 --> 00:02:41.800 The Manifold Cloud is a cloud vertical for MCP. s32 00:02:41.800 --> 00:02:55.360 We provide all the primitives for you to be able to ship MCP servers from your GitHub repo and test them immediately, share them with your team, run emails, run publishing checks to make sure that your app is ready to be put uh submitted, s33 00:02:55.360 --> 00:03:00.040 and many other features that are completely specific to MCP. s34 00:03:00.040 --> 00:03:03.120 Of course, we really believe in MCP. s35 00:03:03.120 --> 00:03:07.720 So, let's look uh a bit in the history of MCP and how did we get there. s36 00:03:07.720 --> 00:03:18.880 So, MCP was launched in 2024, and by the end of uh May 2025, Idel Solomon, the other co-founder, started working on MCPUI, which is this way to kind of MCP servers s37 00:03:18.880 --> 00:03:21.560 that return UI components. s38 00:03:21.560 --> 00:03:30.720 Then, uh this was like many many people started talking about MCPUI because this is like such a great opportunity to ship UI with your MCP servers to agents. s39 00:03:30.720 --> 00:03:37.720 So, it is interactive experiences where the agent is calling tools, but it's showing UIs to the user. s40 00:03:37.720 --> 00:03:39.160 This created a lot of movement. s41 00:03:39.160 --> 00:03:50.640 Many people enjoyed uh this proposal, and the ChatGPT at some point released the app SDK, which is a way to create these interactive MCP apps. s42 00:03:50.640 --> 00:03:54.960 So, basically MCP servers also return UI elements. s43 00:03:55.280 --> 00:04:05.520 Also, at the end of 2025, quite silently, both Character AI and Cloud released and opened their stores for MCP. s44 00:04:05.520 --> 00:04:12.800 These stores allow you to submit your MCP server and have [clears throat] a one-click install experience for your users. s45 00:04:12.800 --> 00:04:24.919 For the most time, these stores were closed in such that were um designed and only allowed for design partners. s46 00:04:24.919 --> 00:04:29.040 But things have changed, and I'm happy to talk about this later. s47 00:04:29.600 --> 00:04:37.480 In January 2026, MCP UI I say converted to MCP apps. s48 00:04:37.480 --> 00:04:46.080 MCP apps is now the official extension of the Model Context Protocol that allows to return UI elements within MCP servers. s49 00:04:46.160 --> 00:04:49.800 So this timeline is kind of an explanation of how the protocol evolved. s50 00:04:49.800 --> 00:04:54.640 And I think still I think two major things happened in this timeline. s51 00:04:54.640 --> 00:04:56.640 First, MCP apps. s52 00:04:56.640 --> 00:04:59.600 MCP servers are not only returning JSON. s53 00:04:59.600 --> 00:05:02.040 And that allows much richer experiences. s54 00:05:02.040 --> 00:05:05.960 And the second thing, maybe even bigger, is that the stores opened. s55 00:05:05.960 --> 00:05:22.080 This was a great uh a huge move by the model providers probably OpenAI, Perplexity, and basically every LLM client out there to say, "Oh, MCP is the way and we [clears throat] want to have a way for people to publish s56 00:05:22.080 --> 00:05:29.360 vetted and quality MCP servers so that people can use them with a one-click install experience." s57 00:05:31.160 --> 00:05:34.080 This is the situation right now with the stores. s58 00:05:34.080 --> 00:05:44.560 The uh that are now being submitted to Character AI and Cloud apps for Character AI and connectors for Cloud increasingly being accepted. s59 00:05:44.560 --> 00:05:56.080 As I was saying in the beginning, this was uh each one of them was gated behind design partnership because the ecosystem was very young, but now it's a GPT and called are both accepting more and more apps. s60 00:05:56.120 --> 00:06:00.720 So, this is a uh this is the moment to publish yours. s61 00:06:01.240 --> 00:06:07.320 So, let's see what MCP apps actually are and what you can do with them. s62 00:06:09.280 --> 00:06:11.640 So, an MCP app works in the in the following way. s63 00:06:11.640 --> 00:06:14.360 Uh much of this is very similar to MCP. s64 00:06:14.360 --> 00:06:20.919 The model is a in the host owns the tools. s65 00:06:20.919 --> 00:06:23.120 The tools are in a MCP server. s66 00:06:23.120 --> 00:06:31.320 But, the MCP server in this case doesn't return a JSON string again, but it returns a widget in a sandboxed iframe. s67 00:06:31.320 --> 00:06:35.200 So, the experience you see is your model is a streaming text. s68 00:06:35.200 --> 00:06:42.320 At some point, it decides to make a tool call, but the tool call is not just returning JSON, it returns a UI just underneath. s69 00:06:42.320 --> 00:06:50.480 And this UI is a sandboxed iframe, so you can as a company developing these apps, you can put almost whatever you want. s70 00:06:50.480 --> 00:06:52.240 But, it doesn't end there. s71 00:06:52.240 --> 00:06:57.200 Actually, there is a bidirectional communication that happens between the iframe and the host application. s72 00:06:57.200 --> 00:07:03.360 So, from the iframe, from the MCP app UI elements, you can send messages back to the host. s73 00:07:03.360 --> 00:07:05.880 You can interact in several ways. s74 00:07:05.880 --> 00:07:09.320 I'm going to talk uh more about that uh later. s75 00:07:09.320 --> 00:07:18.080 The way this works is that MCP uh declares UI resources at initialization time. s76 00:07:18.080 --> 00:07:30.480 When the model calls the tool and it populates the arguments of the tool, uh the tool can then populate the arguments of the UI resource, and that can be then displayed and rendered by the client. s77 00:07:32.040 --> 00:07:36.600 As you see here, this is the kind of experience that you can see in an MCP app. s78 00:07:36.600 --> 00:07:44.640 The MCP server returns tool returns uh UI resource that is populated with the tool arguments. s79 00:07:44.640 --> 00:07:45.320 And here you see. s80 00:07:45.320 --> 00:07:48.360 So, without UI you would see like a wall of text. s81 00:07:48.360 --> 00:07:53.640 The UI allows you to organize the information in a more human readable way. s82 00:07:53.640 --> 00:07:56.840 So, this is uh is a like the basics of MCP ops. s83 00:07:56.840 --> 00:08:01.680 Uh I think they're many many times uh they've been talked about. s84 00:08:01.680 --> 00:08:08.040 And today I wanted to show a bit more of what you can do because this is not often mentioned and I think it's very interesting to design new experiences s85 00:08:08.040 --> 00:08:11.320 which this new protocol allow. s86 00:08:17.760 --> 00:08:27.000 So, first of all, again, the UI is displayed in the chat and it exposes a communication channel between the UI element and the host application. s87 00:08:27.000 --> 00:08:34.039 So, the host application will listen for these messages going through these channels and will react accordingly. s88 00:08:34.039 --> 00:08:36.080 So, this is the first primitive that I'm going to talk about. s89 00:08:36.080 --> 00:08:38.000 So, model context. s90 00:08:38.000 --> 00:08:40.880 In the UI you can show whatever information you want. s91 00:08:40.880 --> 00:08:50.880 For instance, in this case we're showing three articles, but the model doesn't really know it doesn't really cannot really uh introspect in real time what is going on in the UI. s92 00:08:50.880 --> 00:09:01.360 But the protocol mandates this state uh or this set state primitive where you can update the state of the model with respect to the UI components. s93 00:09:01.360 --> 00:09:14.280 So, from the widget itself you can call this is an MCP use syntax that makes this a bit easier, the set state primitive, and you can update what the model knows about what's being displayed. s94 00:09:14.280 --> 00:09:17.600 So, here we have a little demo that shows this. s95 00:09:17.600 --> 00:09:26.200 Of course, uh the the the message uh prompts the tool and the tool shows the UI. s96 00:09:26.200 --> 00:09:33.240 The state uh of the of this UI element is that nothing is selected and the model knows about this. s97 00:09:33.240 --> 00:09:40.960 But if you modify the UI state, you can communicate this state change into the model itself. s98 00:09:40.960 --> 00:09:48.760 So, then basically, if for example, here I send another message, the model will be aware of what happened in the UI element. s99 00:09:48.760 --> 00:09:59.680 And you can do this by simply setting this using this uh um this primitive set state and update the state that the model knows about. s100 00:10:01.240 --> 00:10:02.240 UI message. s101 00:10:02.240 --> 00:10:12.680 So, this is another very cool feature where from the UI element the UI widget that your tool returns, you can send messages back to the model. s102 00:10:12.680 --> 00:10:20.040 So, not only the interaction is I am a user, I have my chat interface, I see the UI, and I want to send another message. s103 00:10:20.040 --> 00:10:23.920 Or maybe there is some contextual message that you can you want to send. s104 00:10:23.920 --> 00:10:30.840 For instance, here we have the same shoe example, and you might want to learn more about Trey Blazer Pro. s105 00:10:30.840 --> 00:10:45.360 And uh you can of course link the click of this button learn more to the primitive send follow-up message, and this will send a message to the chat itself, and the model can start giving you more information about the Trey Blazer Pro shoe in this case. s106 00:10:45.360 --> 00:10:51.240 Um clients have different behavior regarding this, and this is true for many of the MCP features. s107 00:10:51.240 --> 00:11:00.520 For instance, Cloud will display the message in the chat input and tell the user like the user has the choice to send it or not. s108 00:11:00.520 --> 00:11:12.560 While OpenAI is a bit more integrated in this sense, it directly sends the model the message to the model and uh and starts streaming immediately the answer to that message. s109 00:11:17.760 --> 00:11:19.240 This is a very cool feature. s110 00:11:19.240 --> 00:11:26.920 Uh so, again, we have the tool, uh and uh we have a UI element that's populated from the tool itself, right? s111 00:11:26.920 --> 00:11:39.600 If the model streams the input tokens into the tool arguments, you can uh in live take those partial input and update the UI incrementally. s112 00:11:39.600 --> 00:11:51.600 So, as you see in this case, we see that the the the the tool inputs are being populated uh dynamically or gradually, and the the UI reacts accordingly. s113 00:11:51.600 --> 00:12:02.640 I have a very cool demo about this just later in a video where in one of the coolest demos of MCP apps uh in the uses exactly this pattern. s114 00:12:02.640 --> 00:12:12.480 So, you can, for example, imagine like uh you could have a UI component that renders something like an SVG, and there is MCP uh MCP apps doing that. s115 00:12:12.480 --> 00:12:21.720 We also seen uh we actually created a Remotion MCP app where uh we use Remotion to create a video with React, and uh we render the Remotion video inside s116 00:12:21.720 --> 00:12:27.160 the widget in real time as the tokens are streaming in. s117 00:12:27.680 --> 00:12:33.480 Another thing you can do is from the from the widget itself, you can call other tools. s118 00:12:33.480 --> 00:12:44.560 So, first of all, you can call the tool, of course, uh in the in the in the tool that was originally called to gather other data, but from the UI, for example, s119 00:12:44.560 --> 00:12:53.480 you can have a button that triggers another tool call to gather additional data about what what is there in the MCP server. s120 00:12:53.839 --> 00:12:55.760 Again, the primitive is very simple. s121 00:12:55.760 --> 00:13:01.480 This is code on the left here is always um from MCP use. s122 00:13:03.720 --> 00:13:07.240 This is another very interesting thing that you can do with MCP apps. s123 00:13:07.240 --> 00:13:20.480 So, sometimes uh what happens is uh for instance, you uh want your MCP server to return certain informations, but you want to um not show the full information because maybe is uh s124 00:13:20.480 --> 00:13:28.200 there's private information that you don't want to give to the to the model providers, and therefore, [clears throat] you might want to redact that right? s125 00:13:28.200 --> 00:13:34.829 This is like a known privacy issue with MCP servers that you don't want to return and put into the uh s126 00:13:34.829 --> 00:13:36.040 [clears throat] s127 00:13:36.040 --> 00:13:38.040 your private information. s128 00:13:38.040 --> 00:13:41.080 And MCP uh also allow you to do that. s129 00:13:41.080 --> 00:13:46.360 So, when you call a tool, you return a widget, which is populated with some arguments. s130 00:13:46.360 --> 00:13:50.040 You can return other outputs as well. s131 00:13:50.040 --> 00:13:58.760 As in normal MCP servers, the the return of a of an MCP tool is list of outputs of different types. s132 00:13:58.760 --> 00:14:04.480 You can imagine in this case, there is a structured output, which is sent into the widget itself. s133 00:14:04.480 --> 00:14:08.080 And there is an additional output that can be sent directly to the model. s134 00:14:08.080 --> 00:14:15.560 So, something common that you do is you show a very rich UI, like in this case, and uh and this is what the [clears throat] UI will show. s135 00:14:15.560 --> 00:14:23.760 So, this is a card showing the information private information of of this person, but the model will only see the information you want. s136 00:14:23.760 --> 00:14:31.040 So, there's two types of output, the ones that are shown in the UI, to put it simply, and the ones that are sent to the model. s137 00:14:31.040 --> 00:14:42.800 Like a common to maybe understand this better, a common pattern that you see is you show the full information in the UI, and then you instruct the model with a text output s138 00:14:42.800 --> 00:14:45.920 of what the user is seeing. s139 00:14:45.920 --> 00:14:55.560 For instance, we return a uh this this UI card, and we can even return um nothing to the model. s140 00:14:57.680 --> 00:15:03.800 But just say, the user is seeing his private information in the widget above. s141 00:15:03.880 --> 00:15:13.200 So, this is a pattern that allows you to, you know, give uh allow it like uh experiences in in fields where maybe sharing data to the LLM is not possible s142 00:15:13.200 --> 00:15:14.839 because of privacy issues. s143 00:15:14.839 --> 00:15:23.280 In this case, you can show the UI to the user, but the model won't see the data that you display in the UI, unless you choose so. s144 00:15:23.560 --> 00:15:30.110 There's uh another set of functionalities which I think are minor or let's say less s145 00:15:30.110 --> 00:15:32.110 [clears throat] s146 00:15:32.120 --> 00:15:35.080 intuitive or less advanced. s147 00:15:35.080 --> 00:15:44.560 Uh Here we show the request display mode which basically your MCP app widget is displayed in uh in line with the tool call. s148 00:15:44.560 --> 00:15:46.720 But it can even put the full screen. s149 00:15:46.720 --> 00:15:54.960 So the full chat is going to be your MCP widget and uh input box is going to be overlaid on top of the widget. s150 00:15:54.960 --> 00:15:57.360 And for instance, this is very cool for video editing. s151 00:15:57.360 --> 00:16:09.400 You can imagine a a widget showing some uh graphical interface and uh you can chat that we improve the the what's the shown inside the widget and the model can directly stream into the widget you're looking at. s152 00:16:09.400 --> 00:16:14.280 It can also be put in a picture-in-picture or in line which is the normal case. s153 00:16:14.280 --> 00:16:14.651 Um s154 00:16:14.651 --> 00:16:14.680 [clears throat] s155 00:16:14.680 --> 00:16:21.800 there's a other primitives that allow you to open external links from the MCP widget itself. s156 00:16:21.856 --> 00:16:22.320 [clears throat] s157 00:16:22.320 --> 00:16:30.480 You can listen to the theme of the OS so that you know your MCP app is synchronized in theme with the host your users are using. s158 00:16:30.480 --> 00:16:33.200 And there are many other things. s159 00:16:36.040 --> 00:16:45.920 So I wanted to show you here a few videos of MCP apps because so far I've just been showing uh uh uh some mock-up that I created for this presentation. s160 00:16:45.920 --> 00:16:49.480 Uh but this is for example in Cursor we're using the MCP app. s161 00:16:49.480 --> 00:16:53.800 Uh Cursor is one of the clients uh that supports MCP apps. s162 00:16:53.800 --> 00:17:01.839 And as you see here uh our MCP app returns the analytics of the remote MCP server app that I was uh talking to you about just before. s163 00:17:01.839 --> 00:17:06.280 And for instance, this is very uh useful in in analytics. s164 00:17:06.280 --> 00:17:13.760 Uh for instance, we use Pulsar MCP a lot and then the Pulsar MCP will show you a UI element with your with your analytics. s165 00:17:13.760 --> 00:17:22.280 So that you as a human can understand what's going on, but the model itself can read those analytics and go do its job on the code you're you're writing. s166 00:17:22.280 --> 00:17:27.560 Um on Claude, so this is the demo I was telling you about. s167 00:17:27.560 --> 00:17:34.440 So, here we're using the Excalidraw MCP server, and here you will see the streaming functionality that I that I mentioned as before. s168 00:17:34.440 --> 00:17:44.560 So, you can see here that Claude is first reading the some instructions that are returned as tool by the Excalidraw MCP server. s169 00:17:44.560 --> 00:17:53.200 And uh at some point it will call the tool which is showing the canvas, and it will stream tokens into the canvas, and I think we motion it's one of the coolest s170 00:17:53.200 --> 00:17:56.360 animation around how those tokens are shown. s171 00:17:56.360 --> 00:18:05.960 As you see here, this is like a mermaid syntax that is sent into the tool, and the the UI updates as the tokens are streamed in. s172 00:18:05.960 --> 00:18:09.080 There's some of the some of the coolest demos here. s173 00:18:09.080 --> 00:18:13.520 By the way, very very useful to draw diagrams as well. s174 00:18:13.560 --> 00:18:24.360 Um and this is again ChatGPT um uh using the Manifold uh MCP app uh showing the same analytics that I was telling you above. s175 00:18:24.360 --> 00:18:31.560 And as you see here, the rendering is uh is is a bit better in ChatGPT uh it's better. s176 00:18:31.640 --> 00:18:35.320 Basically, this is very similar to how it uses. s177 00:18:35.320 --> 00:18:38.200 Maybe uh this is a good time to talk about the client support. s178 00:18:38.200 --> 00:18:40.400 There's many clients that support MCP app. s179 00:18:40.400 --> 00:18:49.120 Some do more, some do less uh and uh these three I think are the main uh that people are using. s180 00:18:49.120 --> 00:19:02.160 And uh of of course, all the different versions of Claude and ChatGPT uh so both Claude Co-work, Claude Desktop support MCP app, ChatGPT and Codex support MCP app. s181 00:19:02.160 --> 00:19:04.953 Uh and Cursor uh both in the s182 00:19:04.953 --> 00:19:05.160 [clears throat] s183 00:19:05.160 --> 00:19:15.040 agent mode and in the normal side chat supports MCP app, but there is uh many more that support MCP app such as VS code and and and comes as others. s184 00:19:15.040 --> 00:19:29.880 And I think it's actually interesting to to mention here uh that another thing you can do you might be developing your MC server and uh you don't know if the host where your users are using the MC server supports or not MC [clears throat] apps. s185 00:19:29.880 --> 00:19:42.200 What you can do is uh since we know what the client is from the metadata that is exchanged uh uh the MC app uh you can um return a UI element only for those s186 00:19:42.200 --> 00:19:49.040 um host that actually accepts and can render those um those uh widgets. s187 00:19:49.040 --> 00:19:59.240 And then this is not really a big deal because most uh non-MC app clients so uh MC client that don't support MC app will not will simply not show the widget. s188 00:19:59.240 --> 00:20:14.880 But I found developing these servers many times that if you don't show the widget, you need to return a different output because um it some of the information was returning the widget but maybe you want to give it to the model if the widget is not shown. s189 00:20:14.880 --> 00:20:28.120 Uh so this is something also MC user helps you with with some primitives that um that allow you to to know if the client uh your MC server is connected to actually supports MC apps or not. s190 00:20:30.040 --> 00:20:36.480 Again, um a little uh idea of how you can build these MC apps with MC users. s191 00:20:36.480 --> 00:20:40.720 We are one of the most popular SDKs to build these uh MC apps. s192 00:20:40.720 --> 00:20:47.280 The way we design our SDKs that basically you design your MC server uh as you always did. s193 00:20:47.280 --> 00:20:51.880 So you have your MC server constructor and then you define tools. s194 00:20:51.880 --> 00:21:02.600 And from the tools, you can simply return widgets which are automatically registered from this widget file in the resources folder. s195 00:21:02.600 --> 00:21:10.880 So whatever you put as a widget file in the resources folder will be registered as a UI resource that you can return from a tool. s196 00:21:10.880 --> 00:21:20.760 And the widget um the widget file is just a React uh component uh and you can also use your existing uh UI components. s197 00:21:20.760 --> 00:21:25.779 And uh it will be compiled into HTML and CSS. s198 00:21:25.779 --> 00:21:26.120 [clears throat] s199 00:21:26.120 --> 00:21:30.800 And then um returned and linked it to the tool. s200 00:21:31.120 --> 00:21:33.640 We have a scale uh we have a template. s201 00:21:33.640 --> 00:21:39.880 Uh you can just run NPX create MCP app and uh it will give you a template that you can serve. s202 00:21:41.600 --> 00:21:47.280 So, let's talk about distribution and discovery, which I think uh it's uh of course a very important topic. s203 00:21:47.280 --> 00:21:50.880 Uh maybe uh even less known than how MCP apps work. s204 00:21:50.880 --> 00:21:59.840 I'm talking to many people and they don't know there's a store for MCP and they don't know how to submit, so I wanted to talk about a bit about this as well. s205 00:22:01.480 --> 00:22:04.600 So, the store is like a huge new distribution channels. s206 00:22:04.600 --> 00:22:15.080 Again, you're um bringing the three most popular clients, ChatGPT, Claude, and Cursor, which the three of them, they all support a self-serve submission process. s207 00:22:15.080 --> 00:22:17.720 ChatGPT was one of the first supporting this. s208 00:22:17.720 --> 00:22:24.075 Claude, uh since a couple weeks they have um self-serve submission form for team and enterprise uh accounts. s209 00:22:24.075 --> 00:22:41.120 [clears throat] And also Cursor allows you to submit their MCP server set uh The way you submit is different for all three, but basically what happens is that you need to make sure that your MCP app is compliant. s210 00:22:41.120 --> 00:22:49.680 MCP apps or servers can be both uh submitted in all these three stores, so you it doesn't need to return a UI your your server to be eligible for submission. s211 00:22:49.680 --> 00:22:54.720 Um and the three processes to get your app submitted is different and they have different speed. s212 00:22:54.720 --> 00:23:04.160 So, it's going to take maybe a bit more on on Claude uh for now uh and uh ChatGPT instead is it's up a lot the acceptance of these uh of these apps. s213 00:23:04.160 --> 00:23:09.680 And again, the process is you link your remote MCP server. s214 00:23:09.680 --> 00:23:11.360 They will scan the tool. s215 00:23:11.360 --> 00:23:15.200 And they will make sure that all the tools are correctly annotated. s216 00:23:15.200 --> 00:23:22.920 And have the correct arguments and um And once this is done, they're going to scan the the authentication as well. s217 00:23:22.920 --> 00:23:28.720 So, if your server requires authentication, you have to declare it and you need to make sure that it works. s218 00:23:28.720 --> 00:23:33.440 There's a few more different steps which are details for all the all the providers. s219 00:23:33.440 --> 00:23:40.840 Uh but it's important to say that once your app is submitted, it's going to be partially manually or partially automatically tested. s220 00:23:40.840 --> 00:23:44.240 So, you will have to provide some test cases and some test prompts. s221 00:23:44.240 --> 00:23:49.320 And then it will be either accepted or rejected. s222 00:23:49.320 --> 00:24:03.636 And if accepted, you're going to be able to publish it and make it available on the stores that you can find on chadigpt.com/apps on the connectors directory on uh cloud or on the s223 00:24:03.636 --> 00:24:03.920 [clears throat] s224 00:24:03.920 --> 00:24:06.360 the cursor directory. s225 00:24:06.880 --> 00:24:12.600 Again, we we we did the many submissions and we we try to make this process easier. s226 00:24:12.600 --> 00:24:21.800 So, if you want to submit your app, I think you should go to manifester.com where we vet your app to make sure that it's um ready to be submitted. s227 00:24:21.800 --> 00:24:26.880 So, we check and we try to do all the checks that those clients will do in the submission process. s228 00:24:26.880 --> 00:24:39.400 And also, we run we generate some of the submission artifacts that you need to submit like screenshots and test cases for you in in our cloud directory if connected to your MCP server. s229 00:24:40.560 --> 00:24:48.832 Something very very cool about this is that once your app is in the store, not only can people find it by searching on the store, not only s230 00:24:48.832 --> 00:24:48.840 [clears throat] s231 00:24:48.840 --> 00:24:54.360 you can send a URL to your customers and they're going to be able to install your application in one click. s232 00:24:54.360 --> 00:24:59.880 So, you don't have to share that ugly JSON file anymore with your MCP configuration. s233 00:24:59.880 --> 00:25:04.320 But it's very important that dynamic discovery of MCP server is happening. s234 00:25:04.320 --> 00:25:16.200 Today Cloud is the only client that actually does this, but for all apps in the stores, when Cloud needs is like assigned a task that doesn't have a specific tool to do, s235 00:25:16.200 --> 00:25:22.320 it will actually search in the MCP registry for the right connector to do the task. s236 00:25:22.320 --> 00:25:26.280 So imagine what this means for your particular product. s237 00:25:26.280 --> 00:25:42.240 All the people There's many of course active users on those applications or more than billion active users, which will manifest an intent directly in the chat and through the intelligence of the model, the model will choose what is the best connector. s238 00:25:42.240 --> 00:25:56.080 And if you're there, and you do your work to be the connector that is selected, this is going to be like a a huge wave of uh high intent individuals s239 00:25:56.080 --> 00:26:04.760 that want to need your product and will find it uh dynamically and organically on those on those platforms. s240 00:26:04.760 --> 00:26:07.120 So this is very important. s241 00:26:07.120 --> 00:26:14.760 Cloud does this today and [clears throat] ChatGPT is expected to do this pretty soon. s242 00:26:15.560 --> 00:26:19.640 So that was my discrete descriptive part of the talk. s243 00:26:19.640 --> 00:26:29.040 I just want to uh uh I think it should be clear by now how important it is to be on the stores. s244 00:26:29.040 --> 00:26:31.080 I can bring my experience. s245 00:26:31.080 --> 00:26:34.440 Being on the store brought us a lot of traffic. s246 00:26:34.720 --> 00:26:43.680 Personally, as a user of MCP, today I'm checking if a product has an MCP server and that's for me is like the most basic buying decision. s247 00:26:43.680 --> 00:26:55.920 I run most of my day-to-day work on Cloud Co-work or Cloud Code because I love the the the possibility to share the context between my code base and my different connectors. s248 00:26:55.920 --> 00:26:58.120 And this to me is so important. s249 00:26:58.120 --> 00:27:05.080 For instance, one workflow that I that I often run is I have my Granola MCP where I have my meeting notes. s250 00:27:05.080 --> 00:27:07.600 I have Linear where I track my tickets. s251 00:27:07.600 --> 00:27:10.760 Of course, I'm in my code base if I'm using this from Cloud Code. s252 00:27:10.760 --> 00:27:18.640 And I can basically pull the meeting notes with some customer feedback and feed it back in the in Linear maybe I create tickets for the rest of the team. s253 00:27:18.640 --> 00:27:23.320 And then I have the agent that pulls the Linear ticket through the Linear MCP and just starts doing it. s254 00:27:23.320 --> 00:27:24.920 Opens the PR. s255 00:27:24.920 --> 00:27:29.120 And uh and uh and closes the Linear ticket. s256 00:27:29.120 --> 00:27:35.080 And uh I mean, in an ideal world, it would even send an email back through MCP to the customer saying, "Oh, this is fixed." s257 00:27:35.080 --> 00:27:37.240 But maybe we're not there yet. s258 00:27:37.240 --> 00:27:39.800 Um but uh but this is definitely true. s259 00:27:39.800 --> 00:27:47.000 And in fact, just a few days ago, Paul Graham, uh the founder of Y Combinator, said, "Uh AI apps are the new browsers." s260 00:27:47.000 --> 00:27:53.000 And in this case, uh the AI apps are Cloud Code, Codex, Cloud Code Work. s261 00:27:53.000 --> 00:27:55.520 And I fully agree with this. s262 00:27:55.520 --> 00:28:02.120 If you think about it, Google Search in a way has been substituted by looking on ChatGPT. s263 00:28:02.120 --> 00:28:09.264 So, now that we have connectors where you cannot only search, but you can also do stuff in the real world, s264 00:28:09.264 --> 00:28:10.120 [clears throat] s265 00:28:10.120 --> 00:28:15.280 all those all those um operations are going to be moved to the chat as well. s266 00:28:15.280 --> 00:28:22.320 So, in a way, I think uh that if AI apps are the new browsers, the ChatGPTs are the new websites. s267 00:28:22.320 --> 00:28:27.680 And uh as a website, they can return a UI with MCP apps. s268 00:28:28.480 --> 00:28:32.600 So, again, uh I don't want to look at your dashboard anymore. s269 00:28:32.600 --> 00:28:34.800 I want to use it in Cloud. s270 00:28:34.800 --> 00:28:38.280 And if any dashboard, I want to see it in the Cloud Code application. s271 00:28:38.280 --> 00:28:40.200 So, ship an MCP app. s272 00:28:40.200 --> 00:28:40.920 Thank you very much. s273 00:28:40.920 --> 00:28:46.480 I hope you enjoyed the talk and I'm uh hope there is any questions about this and super happy to help. s274 00:28:46.480 --> 00:28:50.880 Um by now, have a very uh expert on the topic. s275 00:28:50.880 --> 00:28:52.360 So, thank you very much. s276 00:28:52.360 --> 00:28:54.760 Have a good one.