Real-Life AI Content Workflows with Directus MCP Workshop

Speaker 0: We are live.

Speaker 1: Hang on. I got something for this. Sorry.

Speaker 0: Alright. Let's let's kick this thing off, baby.

Speaker 1: Nice. Got a lot of folks in here.

Speaker 0: Alright. Well, Matt, you wanna do the honors? You want me to? What do we how are we kicking this thing off?

Speaker 1: Oh, yeah. Alright. We're going straight into it. Love it. Yeah.

So hey, everybody. Welcome to our, Real Live Content AI workflows, directed some TP event. My name is Matt. Bryant is sitting in the copilot chair. I guess you're more of the pilot because you're controlling really everything.

I'm just gonna sit here. But, yeah, great to see everybody. You might be seeing a couple people drop in where they're from. That's kind of the classic session workshop standby to open things. So cool to see, like, international representation in there.

Cool. You wanna kick things off, Brian? Yeah. Yeah. Yeah.

Yeah.

Speaker 0: So, over the last, I don't know. It's been six weeks at this point. We've, like, been working on this MCP. This is the launch event for that. But as we kick things off, definitely let us know in the chat who you are, where you're from, what you're building.

We've got, like, a a few short slides we're gonna dive into, and then we will showcase these actual workflows that you can take and use right away. So I'm super excited for that. I'm sure we'll forget something along the way, but we'll have some fun with this. And, the whole goal here is to, again, give you value that you could put to use right away right after this call. So, Matt, with that?

Yeah. Alrighty, man.

Speaker 1: Let's roll. Alright. Well, like I said, good morning, good afternoon, good evening, everybody, where you're from. My name is Matt. I've I'm obviously from Directus.

I'm on the marketing team here, so I head up a lot of the content initiatives and that sort of thing. So primarily the non technical and non doc stuff that you'll find on our blog and our case studies and all that stuff. So I figured it would make sense for me to kind of kick this off because content is near and dear to my heart. I've been doing it my entire career and I know the pains and struggles of it throughout my entire career too. So just on the agenda today, just so you know what we're going to cover, we did the introduction, so checkbox there.

Hopefully, you know who we are by now. I'm going to talk a little bit about kind of the bigger problem of what the MCP actually solves and kind of how we're thinking about AI, generally speaking there. Collaboration, beyond just the content. So this is a very like small microcosm use case. So I just want to set the scene correctly.

So you understand like what the MCP is, what it can actually solve for your business, or your, your clients, and then we'll hop into the live demo. So I'll get all the boring stuff out of the way first, with the story typical marketer. And then we'll hop into, like, kind of a q and a. So to kick things off, when I think about content, I always think of the hours I lose on kind of the stupid mundane tasks. And I saw a stat that content teams only spend about 45% of their time on real value generating work for the business, right?

The rest that remaining 55% is consumed by the operational administrative tasks and that sort of thing. That's how much time the average content content team spends on work that isn't actually creating content, right? It's like copy and pasting from Google Docs, fixing broken formatting, building page structures manually, tagging assets one by one, and then losing all the assets. 55% of your team's creative energy goes to busy work. Imagine if they spent that time creating, not copy and pasting.

Trust me when I say that I've tried like every AI out there from cursor to Jasper in the early days to Claude perplexity. Now you name it. I've tried it. Most AI tools out there today will help you write a better headline or generate some copy. And that's useful, right?

That's good. But it's not really transformative. You still have to move that content. You still got to format it. You still got to publish it.

You still got to manage it. That's the 55% we're talking about. So what we built with this MCP is different. It's not just another GPT wrapper. Instead of AI that suggests what the headline or what the content should be, we decided to build this MCP to connect it with your CMS and handle all of that busy work as well.

So Yeah, in the next forty five minutes you're gonna see content move from Google docs to a live site without touching copy paste landing pages through conversation, analyze and organize which libraries automatically. And I just wanna stress, like, this isn't a prototype or a concept. Like, it exists right now and you're gonna get access to it after this. So, yeah, it's the direct SMTP. We're excited to give everybody access to it.

We just got to get through kind of the live demo just to show you how how it works first. Brian, if you can scroll up just a few to the

Speaker 0: This one? Yeah. Down. Okay.

Speaker 1: It said before we dive in, I just wanna zoom out for so what you're about to see with the content team is really just the beginning. Right? Like all of those mundane tasks that 55% that is lost. Content is just a perfect microcosm where AI is heading across every company. It's the same problems the content teams face face I know firsthand, that really everybody faces.

Right? It's disconnected tools. It's the AI that suggests but doesn't execute. It's happening everywhere. I think we got a few sales folks on here too.

Sales teams are copying new data from forms and other CRM, sending the email tools, and then the reporting dashboards, same date. You got developers. I think most of you writing the same credit operations, right? Like building the same admin interfaces, creating the same API endpoints, handling the same data validation logic on every single project and then maintaining it all separately. And then you got leaders are asking like, what's the status, from their teams and getting different answers because everyone's working in different systems that don't really talk to each other.

So all of that to say, content teams aren't unique in these struggles. They're just the clearest example of a universal problem. And what you're about to see really is a glimpse of how that whole disconnect can be solved, not just for content teams, but you know, the context made sense. So zooming in again. So most AI tools you probably know today are, like, having a really smart intern or a really dumb intern, depending on how you use it, who could only give advice.

Like, here's what you should do. But then you still have to actually go and do it. The real revolution here, and I hate the term revolution, I guess, evolution is that AI that actually does the work right? Ai that connects to your systems, understands your processes and executes within your business rules. You've probably heard about AI agents, right?

If you have, this sounds pretty familiar territory to you. If you have it, it's basically autonomous AI that can plan and execute complex tasks across multiple systems. Like that's exactly like exactly where the industry, every industry, all businesses heading. But here's what most people miss is that agents actually need infrastructure to work. They need secure access to the data.

They need permission systems. They need to understand your business rules and they need APIs that actually connect to your workflows. Most companies are trying to build agents on top of disconnected systems and that's like building a highway with no on ramps. What we're going to show you today, the MCP connecting AI to your content workflows. This is really the foundation for what makes AI agents possible.

You can't have intelligent automation without intelligent infrastructure. And I just wanna caveat this first by saying, like, we're not at all advocating to replace humans. I think one of our competitors came out yesterday and was talking about like automating 10% of The US labor force. Like it's kind of messed up. I don't want to talk about that.

I hate when people talk about AI taking people's jobs. That's not what the focus is. The focus is really creating a true collaboration between your team and the AI so they can be better and 10 x their output. You bring the strategy, creativity, and judgment and let the AI handle all the execution and consistency and scale. So, when that collaboration works, you don't really just get productivity gains.

You get entirely new capabilities, right? It takes your business to the next level. Everybody wants to be valuable to the business, and this is the clearest way cutting out that mundane work and focusing on strategy and creativity. I probably sound like I'm two X right now. I think somebody mentioned with my, talking, but, yeah.

All right. So to wrap it up, just this concept of the collaboration layer is, this is what's stopping most things now. Companies build AI features inside of existing tools and we think the future really is like kind of the the infrastructure outside of it. It's the collaboration layer that sits between AI and all your business systems. That's what this MCP is like.

You've got your presentation layer, your front end, your websites, your apps on top, and you've got kind of the bottom foundational thing, database back end systems. But what's missing is this middle layer, the collaboration layer. It's kind of the connective tissue that lets AI work with your CMS and your database and your workflows and your team processes all through one interface. That's what direct is sort of becoming, you know, it's not just a CMS with AI features. It's a collaboration layer that lets AI work with your tech stack.

Content management really is just a first use case and a lot of people use it for that. And it's really good at that. And the architecture we're building, I think, works for any workflow, any team, any business process. So I just want to emphasize, like, this is a small glimpse into what's possible when you start using this MCP. And I think, you know, it's, it's massive for, and I hopefully like you leave this with some ideas of like, this is applicable for a lot of different use cases and, for my clients and business and that sort of thing.

Speaker 0: Cool.

Speaker 1: That's the future. So I'll, I'll stop talking, get all the business stuff out of the way. Ryan, I'd love for you to take over and actually show them, show the final couple.

Speaker 0: Show show the people. No. Hey. Hey. You they perfectly said that.

Again, like, to me, the promise of AI is is not about taking jobs or automating 10% of the labor force. Right? It's automating the stuff that sucks, like the headaches, the stresses, all of that stuff. Alright. So before we dive in, let's just cover, like, a couple of MCP concepts just like, I'm sure a lot of you are already familiar with this.

But, for those who are are not super familiar, I haven't dove into the deep end on this. MCP is just a a standard that is, becoming widely adopted or is already widely adopted for helping AI tools communicate with, your different applications and services that you use, like Directus. So model context protocol, and that's the long form, basically just sits in the middle, and it standardizes the way you communicate with LLMs from all the different services. We won't go into the nitty gritty on that. That's not why you're here.

So what is the direct Us content MCP, and why is it the Content MCP? So this release is mostly focused on working with your existing data inside your Directus instance. So it it a lot of it is centered around content. Like, the tools can be used to publish blog posts, fix links, build landing pages, organize assets, update menus, and I'm sure you'll come up with a thousand other use cases for this after this call. But it does not include potentially destructive things like, deleting tables or collections from Directus.

Potentially looking at that in future releases, but, mostly, we wanna give you the tools to work with your existing data and not potentially blow up your Directus instance. Alright. Cool. Everybody still with me? Alright.

So how do we get access to this thing? How do you work with it? Depends on the LLM or the AI tools that you're using. There are three ones that are pretty popular right now. Raycast is, kind of a command bar tool that you could see here on my screen.

But for this demo, we're gonna be using Cloud Desktop. It is the best one that I've found for these content workflows. So, we've got documentation. We've got all of that you can take a look at after this, but let's just take a look at how you install this thing and work with it. Alright.

So the first thing you'll need, you'll need to download Cloud Desktop if you don't already have this. ChatGPT does not have MCP support yet, but some of their APIs do. So download this, and then there is a configuration that you'll need to do. So we'll open up Claude. We'll go to the settings.

You'll see something like this. You click over to the developer tab, and then you're gonna click this edit config. Now, for any non developers here, this might look scary, but, let me just comment out my token here. We're just gonna pop in some JSON config. The two pieces that we really need here are the directus URL and a directus token, which you can grab from your directus instance.

If you don't have a Directus instance, would love for you to try us out on Directus cloud or, you know, we'll give you the resources that you need to set up a a self hosted instance of this exact setup, at the end of this thing. So this is just the command to connect to the Directus MCP server. Right now, this uses the standard IO format. So, if you do any research on MCP, you'll see that there's a a lot of confusion on security and privacy, and there's a lot of different attacks. Right now, this runs the MCP server locally on your computer.

So, we circumvent a lot of those things. And one of the beautiful things is this respects all the permissions that you have inside Directus. Alright. So, onto the fun stuff. And I'm sure Matt, I think he's did you mute yourself, Matt?

Yeah. Probably so. Anyway, this is this is a workflow that you know well. Right, my friend? Publishing content.

And just about every content workflow. Right? This is how it ends up. So you know that all roads start and intersect through Google Docs at some point. We've got articles that we want to publish here.

The suggest, edit, all that functionality that's built into Google Docs makes this hey. It's just a easy tool. Right? And Directus is about working with the tools that you've got and instead of trying to force you into editing your blog posts or content outside of that. So the first workflow that we're gonna take a look at is how to get this blog post that we've got into our direct us CMS.

Alright. So here's our blog post. The standard format looks like this. We go in. We open up a new blog post, and then I'm immediately greeted by a lot of different fields that I have to fill out to make sure this gets published correctly.

And then you play this game where we copy paste into the rich text editor, and you'll see that Google Docs has all this formatting and all this extra stuff that I'm now gonna have to go through and and account for. This is gonna eat up a lot of time. Not only that, like our SEO meta descriptions, the short summaries, the content writer left all that out of the Google Doc conveniently. So now I've got to go in and write 60 characters that make make this make sense in a search pane. Let's let's just make that workflow a lot easier though.

Right? So we are going to go into Claude. Once you've got that Directus MCP server connected, you'll see the available tools here, and I, you know, will explain some of these as we go through. But I'm just going to add this article from Google Docs. So this is a built in integration, and we're gonna say, hey.

Please, always say please, add this article as a draft post inside Directus. Alright. So once I do that, what is going to happen behind the scenes? Right? This MCP server is going to connect to my direct instance, and you could see Claude here is calling the system prompt that we've got set up for it, which is basically some instructions of how to interact with the CMS.

It is going to pick up the exact schema from my direct assistance. So this is, understanding all of the individual fields that I've got set up inside my direct Us instance, the data model, all of that so that the LLM has the proper context to do this job for me. So now you can see it's figuring out there's a post collection. It is automatically written a description for this piece of content, and you could see it's currently in the process of taking all of that ugly Google Docs formatting and just doing standard HTML elements that are going to look on brand, look amazing inside our, website. So this has finished this workflow.

I can just click on, that. Or if I refresh over here inside Directus, Now we can see we've got our blog post. Right? And I can show this in the live preview here. I can also open this up in our visual editor.

So this is that collaboration that we're talking about where AI gets me most of the way, and then I can go in and tweak things here, like, visually inside Directus, or through the the form space approach. Right? Now, the next workflow that I wanna show you guys. Right? This is pretty impressive.

Feels like we're getting closer to the promise of AI. But I want to talk about assets for a moment. So this blog post needs a featured image. Right? And any content editor, any any CMS user has always been is is a % been through this pain of a file library that looks like this, image dot PNG, image zero zero zero one, random UUIDs.

Right? So let's just take a look. I'm gonna go to Unsplash, and I'm just gonna download some random images here. So this is a sunset. Great.

There's a penguin. Not sure if John from our team is on this. I know he's a big fan of penguins. And we'll just grab one more image. Right?

So, also going to showcase the ability to just import from a URL inside Directus, but, also just gonna drop these images that we've got. We've got them here. Cool. We've got these images, but no metadata. Like, if someone is looking for an image of a penguin inside our CMS, So if I just go to my file library and I search penguin, oh, it looks like I've got a penguin already in here.

But I digress. Right? This is a problem with every CMS. Now, what I can do with this MCP is automatically organize these assets. And I can't overstate how huge this is, just as a person who has a ton of unorganized assets in the CMS.

Alright. So two capabilities we're gonna showcase here, the ability to load props inside, something like Cloud Desktop. So I've got this prompt where it is, organize images, and I'm gonna tell it to organize all the images in the needs organized folder. Alright. So I'm gonna add this prompt.

Where does this come from? What what is happening behind the scenes? Right? So inside Directus, we've got just a AI prompts collection, and I could store these to allow all the users of this MCP server or anybody who has access to it to recall these prompts and, build really robust workflows that are super easy to use. So, basically, it's just taking this.

It's added some of the context. These are instructions we're gonna use over and over again. I'm going to send this, and the Directus Content MCP server should go about organizing these assets for us. So it is going to, look for a folder that has needs organized, then it will read all the files within the folder. And one of the tools that we have available is the ability to read the raw image content.

So it's actually looking at the the binary image data using the vision tools inside Claude, and actually analyzing that image. And it's letting us know, yes. Hey. I've analyzed all three of these images. Let me update these for you.

The king penguin portrait. Great. And we could see it has the title. It has the description. It has the tags.

And one of the other nice things, especially for a front end, Directus has support for focal points. So whenever an image gets cropped or transformed dynamically, we wanna make sure that the the subject, the the focus stays in frame. Alright? So we can see here that Directus has organized all these assets for us, and now we are able to search these and and use these in our our blog posts, in our workflows. Alright.

Next one. It's hard doing this with with, nobody else on the on the line here, but maybe we cue that laugh track at some point. But, next on the list. Right? Building landing pages.

And I say this a lot if you've ever caught any of my YouTube videos. One of my favorite features of Directus is the ability to build dynamic pages using blocks. So if you're a developer, you could set up different blocks for marketers, for content editors, and they can dynamically build pages based on those blocks. With the recent visual editor release, I can control these blocks dynamically, edit them right in line. Huge.

But scaffolding out these blocks is still fairly tedious. Right? It is a a lot of forms to fill out, and just a a lot of back and forth to to get to something where I can tweak to to what I need it to be. So I'm just gonna start a new chat here, and let's do something more complex. Right?

So we're gonna build a landing page. And I'm a little embarrassed that I didn't prepare for this one. I haven't had time to actually run this prompt through the system. It is, one I created right before this thing. So we'll see how it goes.

Hopefully, it has fun. But just a a costar prompt that says, okay. Here's how I want you to structure the landing page, what goes into it. And I am going to call this from the cloud desktop. So we got create landing page, and we're gonna do this for direct to CMS.

So direct to CMS desired action. We want someone to sign up for a cloud trial. The value proposition, a better CMS, headless, not brainless. That's my key value proposition. Who is our audience?

Right? We want, technical marketers and developers. Ship faster is the specific pain point that we're solving. The specific solution and outcome is better workflow, if I can actually spell, less stress, cost, time to implement, migration. Alright.

Cool. So now I've got my prompt finished. We're just gonna add this to our workflow. I'm gonna jam this into Claude. And, it again, it's gonna go through this progression where it understands the role that we've assigned it.

It's going to pick up the direct to schema. And now it knows the available blocks that it can use to build a page for us. Right? So it is going to go through this whole progression of, creating a page. So we could see the page that it has created here.

No content in that yet. Right? So it's actually going through and systematically creating each one of these individual blocks for the page. I can now go away, have a coffee, walk the dog, come back in just a minute here, and have a fully scaffolded out landing page for me to customize and tweak and and get ready to publish. Alright.

Gotta get that product placement in there. So we can see this thing is working through. And, you know, obviously, like, the the benefit here, again, like, it knows your direct to schema. And so, you know, you can get access to this CMS setup that we've got. I'll just, go to the website here while it's we're waiting on this to do its thing.

Directus.io. If you're a developer, you can pull this down. If you're not a developer, just click get started free, and you can actually get this exact template. But if you are not operating with this exact template, right, it pulls that schema and data model from your instance so it knows exactly what the fields that you have are and what belongs in that content. Alright.

So, hopefully, this thing will get finished in just one moment. Let's take a look at some of the questions. Do we have any? Mister Matt, do we have any major questions while we're we're waiting on our friend to to do his work here?

Speaker 1: No. We got some good ones. Yeah. Keep the questions and stuff coming in. We'll we'll try to get most of them, at the end.

But, yeah, I was just responding to Kent who's got a deadline tomorrow to convince his boss to create new blocks. So, Kent, if you need, like, a doctor's note, we'll, we'll go ahead and handle that for you. Juan had a good point about, you know, the the user experience is really nice, but, what about the time it takes for a large inventory? Right? Like, lots of files, lot of lot of images when you're going through and, like, doing the organization and that sort of thing.

Does it scale up if needed? I think, you know, it depends on the LLM or the model you're using. Obviously, like, the new ones, like, the deep thinking ones will take a lot longer, but, on your, you know, early stuff, it could be quicker.

Speaker 0: Yeah. Certainly. Like, you could, I you know, like, Claude Haiku, I think, is, like, their fastest model. I believe it has vision. Don't quote me on that.

It just basically depends on the the actual model that you're working with. You know, I'm using, like, the latest and greatest from Claude here because I I'm not only asking it to publish the content. I'm asking it to, you know, do some massaging and and actually help prepare the content in our voice. And I found that Claude's on it is just hands down, like, the best rider for that. Again, like, this is the MCP server is just a protocol for connecting to different AI tools, and Cloud Desktop is is just the use case we're featuring here.

But you could certainly, connect to Cursor or Raycast or, any other tools that support this protocol.

Speaker 1: Yeah.

Speaker 0: Alright. So it looks like it has done the work here, and we've got our landing page ready to go. So let's just open this thing up. We'll take a look. We could see we've got a full landing page here.

I'm just gonna open this up in the visual editor. And now we can go through this thing and see that, hey. This is got everything scaffolded out for us, you know, really nicely. Right? And I'm in a position now where I can go back in and just quickly edit whatever I need here.

Hit save. You see that change and update.

Speaker 1: I I can't stress, like, how much of a game changer this is. Like, you know, it typically takes days just for the scaffolding and there's, like, you know, writer's block and you just don't know what blocks to put. And this one is just, like, bam. Like, it's all there. And then when you combine it with the visual editing and hand it off to the content team, it's

Speaker 0: again, collaboration. Right? So using the AI tools, also being able to visually edit and quickly tweak that. You know, this is not about completely eliminating. It is about streamlining, about taking the stress out of it, about getting something that I can then take and improve.

The last but not least, workflow that I wanna show here is, publishing and automations. So, I love this meme, but picture automated AI agents, in this case, Claude, that are calling automated workflows that you have set up inside Directus. So Directus supports, low code, no code automations depending on what you're trying to do. Is a feature is called flows. So it is a visual automation builder, and I could do complex stuff.

I could do really simple stuff. In this case, let's take a look and hey. This looks great. This looks great. Change it to published, and then publish the entire site using flows.

Alright. And and just say, like, best practices here. Obviously, like, the better you prompts that you have, like, what you're putting in, the better results you're getting out. And I'm basically playing, dumb here a bit and giving it, like, simple instructions. But, obviously, the the more detailed you can be, the better the results.

So what's happening here? Right? It is updated the status of this page and added the the current date and time stamp for me. Page is now published. But the next step here, you could see that it is reading through the flows.

It has found a flow called publish website, and it's actually triggered this flow. So if I go over to our, Vercel account, you can see that this is actually going through and rebuilding the site. And, you know, in fifty four seconds, this content will be live on on the main site.

Speaker 1: Wow.

Speaker 0: And, and, Vercel trying to get the the extra the dough there, upgrade to pro. Alright. So that is the demo portion of this thing. You know, we have it basically speed ran this thing, I think. I hope it is, it just said, like, a good showcase of of what's available through the ContentMCP.

There are, like, more tools in this, that are available. One of like, I hope this serves as inspiration to you all of what you can actually do with this thing. And I can't wait to see the workflows that everybody cooks up for the MCP. Definitely share with us. Keep the feedback coming.

I that's all. That's all I've got. Matt.

Speaker 1: That's awesome.

Speaker 0: Back to you.

Speaker 1: Yeah. I've I've I love watching it every time. Cool. I'm just kinda running through some of the, the comments. There's a lot of stuff, a lot of questions around kind of the costs credit situation when you start using obviously, when you start using, like, LOMs with yourself and especially if you give access to, like, the marketing team, for instance, who tend to run through a lot of credits, not speaking from personal experience.

But, anything on that that just, like, you know, in terms of maybe that landing page. Right? Like, how many credits or whatever the cost could be for that?

Speaker 0: So as as far as the credits, like, the demo that I've got here is using, like, a I've got a paid subscription to Cloud Desktop. So, you know, I'm sure somebody in Tropic is maybe screaming about, you know, especially all the testing I've been doing over the last couple weeks, especially on some of the the top tier models, but that's included in the subscription. You know, depending on which models you're using, it's not really much different than than using these via API or whatever. Like, you know, there's always a balance that you try to strike between how effective it is and how, like, how much you're actually spending. And I you know, the same balance here.

I don't I I don't see any any, like, specific stuff that you wouldn't encounter using AI in in in the same fashion.

Speaker 1: Yeah. Yeah. I think, yeah, we've we've never run into any credits issues with this stuff. But actually, we've been talking with Century. If you're familiar who's actually piloting they've got a beta out, I think, of a, it's like a LLM monitoring, so they can track, like, how many credits are being used.

So it could be a good thing to test with this. We're currently talking to them about, like, a potential, you know, co co workshop for something on that. So we'll definitely keep everyone posted if that moves forward. But that could be a great thing to pair with with this so, you know, you're not going, bananas with your credits. Cool.

Speaker 0: So one other question I saw here. Does the MCP support having multiple direct Us environments? Like, one instance of the MCP does not support multiple environments at at this point. Right? But there's nothing stopping you from potentially, and I I don't I'm not sure that I've actually tested this, to be honest.

But, you know, I don't think there's anything stopping you from trying this out where I could say, hey. This is Directus non prod, something like this, and creating multiple instances of the server. I just worry about it getting confused depending on, like, the model and the tool that you're using. So try it out, report back, see what you come up with. It's the best advice I could give.

Speaker 1: Paul, shout out Ridiculous Engineering. How's it going, Paul? Good to see you. This is actually a great idea, about, you know, some sort of exchange, MCP kinda workflows, prompts, marketplace. What do you think, Brian?

Speaker 0: I would love to see that. One thing that we're gonna do is, like, all this, say, like, the slide deck, this FigJam, like, all this, we're gonna send to everybody who's registered for this event. We'll make sure you get access to that. All the prompts that I've used today, we'll include those. I you know, for now, I would say we're just gonna direct those to the community.

So if you haven't already, sign up for our community. It's just community.directus.io. We've got an MCP server thread. You know, maybe we start a thread on MCP workflows and prompts so that that we could share and, you know, prosper together. Like, everybody take advantage of of AI.

Right?

Speaker 1: Let me just drop that that here. So, this this you can get access to it, on our new community platform here. Right? You were just saying? We've got it linked.

Speaker 0: Amazing. Yep.

Speaker 1: Cool.

Speaker 0: What else do we have?

Speaker 1: Here you go. Nice little shout out. No explosion. For the demos in in one.

Speaker 0: It's Obviously smoothly. I love it, man. Obviously, George has been on some of my other demos before where I've got, like, multiple Docker containers running, you know, like, three different Nuxt applications and, yeah. No. This is a good one.

Thank you. Appreciate you, George. What else? I I see this one's from Jonathan, but it is, Jonathan's on our team, the magic man. But, I probably in response to another question, the current MCP release.

Again, if we just take a look. MCP. I'll share my screen again. This is the actual repository for it. It just got, like, a list of tools somewhere on here where we can just quickly run through these things.

Like, we can read the current user information, read collections, schema, create items, update items, read files, import files. So all this to say, like, all the admin functions that are available through the Directus API, we didn't make available in this release. We wanted to make this one very focused on working with content inside your Directus instance. And, you know, Directus has got robust permissions so that, you know, a content editor who doesn't have access to collection a can't do that through the MCP. But also one of the concerns with production data is security not blowing up, and then since not accidentally dropping a table from the database and losing a 50 pieces of content or 3,000 pieces of content.

So we didn't include those more destructive tools in this release for that very reason. Now we are exploring what that looks like in future releases and how to add some more of those powerful tools, especially if you're, let's say, building a new instance for a client or for an internal team of, you know, maybe modeling out some of your schema for you, like, creating relationships, configuring permissions, all that stuff that is is necessary. But, you know, maybe not, like, the best use of your time to a point and click and set all these things up.

Speaker 1: Thanks. Here's one.

Speaker 0: Oh, I killed it. Sorry. Just trying to

Speaker 1: creating collections on the

Speaker 0: Yeah. There there it is. Yeah. That's the exact question. Yeah.

So right now, no. It cannot create collections at this point.

Speaker 1: Damn. Alright. Just kinda scrolling through.

Speaker 0: Anything else?

Speaker 1: Soft deletes would be great.

Speaker 0: Soft deletes would be great. Yes. And that is like a per collection setting inside your direct to CMS where you can archive. I I don't think I've got that set up on this demo instance, but it's totally doable.

Speaker 1: Cool.

Speaker 0: Any other final questions we need to answer?

Speaker 1: I think there was one earlier about, you know, can this MCP be used to create extensions or kind of start to kind of expanding the use case out of just, like, kind of the the rudimentary content stuff?

Speaker 0: Yeah. Yeah. So certainly, you know, again, like, looking at a a future release where we have some more available tools, like, a builder MCP, if we're calling this content MCP, don't don't hold me to any of those things. Still experimenting with a lot of that. There's a lot of promise.

I will say that. But this is, again, this is mostly focused on content teams. Depending on what content you've got in Directus, you know, there's a lot of possibilities there. But if you're a developer working with Directus, you're setting up Directus for clients, just stay tuned. We've got we've got some fun stuff coming for you.

I promise.

Speaker 1: Cool. Alright. Well, I guess we can wrap up early. I do wanna pull up one thing. If you could just pull up that, photo, like, the image, asset manager from the direct to search lens.

If you've got quick access.

Speaker 0: So the actual Yes. Like instance? Yeah.

Speaker 1: Within the Directus instance. And just, make that penguin picture bigger. Bam. Alright. Michael said that's one judgmental penguin and, like, I just glanced at it.

And I was like, yeah. That's the most judgmental penguin I've ever seen. So, yeah. Just wanted to call that out. And I don't know how to know.

You know?

Speaker 0: Yeah. Alright, guys. Matt, thanks for putting together this event, doing the outline. For everybody in the audience, thanks for attending. Like, again, I run with this thing, build cool stuff.

Make sure you share with us. Like, everything that we build is a % for you guys. Like, we wanna solve real problems, with these tools and and not automate 10% of The US labor force. Again, to get access to this, you can go to directus.i0/docs. There's an AI section in our documentation.

Again, we'll send out links to all this. It will give you installation instructions. There's videos of of me walking you through that installation process. You'll be freaking tired of my voice, by the time you're through. But then we also have, just sample workflows that we showed today, and information about the other configurations, like setting up the prompt system and all the other good goodies that are in here.

And if you want the simple CMS template, again, get started free. Choose the CMS template when you go through the cloud creation process. Or if you want that locally, just copy this command, run that, and you'll be up and running in less than five minutes.

Speaker 1: Cool. Yeah. And like Brian mentioned, we'll be sending out the, attendee resource kit with links to the repo, community.directus.io, and everything else you need to really get started on this today. So, yeah, I'm excited to see what comes of it. And I love that idea of a marketplace potentially.

So, we'll definitely have chats about that. But yeah, thanks. Thanks everybody. Appreciate the time today.

Speaker 0: We'll catch you on the next one.