See the Directus MCP Server in action as we document the actual workflow used to create this entire series. Watch real-time AI collaboration transform a simple list of videos into a complete, structured content series with full metadata, relationships, and automation.
Speaker 0: Hey, guys. Brian here for Directus. And in this video, I'm gonna give you, like, a behind the scenes demo of the Directus content MCP. Can't stress this enough. We just wrapped our MCP event.
Same shirt, same same hat, all that. This is not scripted. I've not practiced any of this stuff. You know, so many of the AI demos and tools that I see are these polished scripted videos that have been practiced to the gills and it's hey, here's what you could do with the tools. Look at this.
You can potentially do this. I'm gonna show you something that you could do right now with the Directus Content MCP and go. Okay. So, we have a platform called Directus TV where we, have a a different series of shows on all sorts of different topics inside Directus, like translations or mapping or authentication or theming if you're into that. Don't know who that guy is.
But, we put all the recordings of our workshops into, something called enter the workshop. But this Directus MCP, we've got seven or eight videos. We're gonna create a new series. So this is behind the scenes. This is our actual instance.
I'm gonna set this up. I'm gonna go in. Just gonna log in to my account, create a token. Cool. I'm gonna hop over to my MCP server setup, and I'm gonna plug in the TV URL.
Hopefully, I blur this out. And I'm gonna take my token. So I'm just gonna paste that here, quickly move that off screen. Obviously, gonna have to roll that one after this video. And then once I'm done with that, I'm just gonna restart claw desktop.
Awesome. Alright. And can you check my directus schema for me? Now I'll be super happy when Claude desktop supports, like a voice mail or something. But let's see what we've got here.
So it's loading our system prompt. Cool. Let me examine your direct to schema. Sending a request. Appears to be a content management platform for TV video platform with episodes.
Awesome. Alright. So now, like, how to go about this. Right? Here's our organization structure.
We have a show. Each show has seasons. And within each season, there are episodes. So I'm just going to try to do this in the most lazy possible way. How can I copy all these names of the videos?
Direct us NCP. I want to create a new show for the Directus NCP server. Please create the show, and I'll send you the links to the videos. So there are videos that are hosted at Vimeo. And if we take a look at the actual structure and you could get this exact same setup inside Directus plus.
But you can see here's the structure. We've got, a season. There's an episode number. There's a slug. There's a title.
There's a Vimeo ID, and that's how we store this. And then we have episode people and lots of kind of complex relationships. There's also flows that are set up to generate transcripts. So, Vimeo videos. Alright.
So I'll just give it more context here. We're gonna say, hey. Here's the ones I have. I have, Claude desktop MCP installation, cursor installation, Raycast installation. What do we have?
We have a blog post workflow. Workflow. Real life AI content workflows workshop. And what? Landing page workflow.
And what's the last one? Organize assets. Alright. So I don't know how well this is gonna do. I'll just copy paste the videos.
Link. That's called desktop installation. This is cursor MCP installation, Raycast installation, blog post workflow, landing page workflow. And if I had to do this a lot, I would obviously try to optimize this even further. Landing page workflow, real life content workflows.
Okay. Alright. So I'm not telling it anything about the structure. Right? It's already picked up the schema.
And let's give it some more context. Here's the docs for the MCP server for additional context. Alright. So I'm gonna just go here to directus.i0/docs. And there's, like, a a lot of descriptions and and things like that that go into this.
I'm just gonna paste these docs, tools, prompts, workflows. Alright. So that's all the docs that I've got. I'm just gonna trust and and fire this off that in the episodes that this is actually gonna work out based on info below. Now look, I'm, again, not a prompt genius.
This is all stuff that you can do right now. And if you learn a little bit about prompting more than me, you will get even further than this. So let's just see what happens. Is going to create the show for me. There's the description.
This is gonna be published. It is a AI powered content management with Directus and MCP. Alright. So now it's creating a new season for that specific show. Looks like it is creating the individual episodes.
Alright. So now if I go in here, we search for MCP. Right? I can see this. There's the card text, published, status, images.
So we we need some images. I don't have to generate those, outside of this workflow. But if I go in, we look at the season, I could see that it's starting to populate the episodes for me. And is it done? Let's create the workflow episodes.
I don't know if it was actually done or not. Alright. Organize assets. The workshop episode. Organize assets.
Workflow. I'll just save and update that. So, again, hey. Like, on this workshop, we covered this collaboration model where, you you know, hey. I'm actually working alongside this, and I I think this is a great just a great showcase of that.
Right? This is the MCP server. Now all I need to do is basically just test this out. Right? I've got the show.
I've got the season for the show. I've got the episodes for the show, and all the descriptions are there. It looks like it's missing the episode people. So can you add what am I set up as? Is there Brian in here?
There's also Brian Ross. I don't know who that guy is. Can you add Bryant Gillespie to the episode people for all the episodes and add Matt Miner for the last one, the workshop. Alright. So this is, you know, a fairly complex relationship.
Like, this is a many to many relationship. And okay. It should so it's going to go in and query all the TV people to see if they exist. Then it's gonna go into add Brian Gillespie to all seven episodes and Matt Miner to the workshop episode. Now, again, this wouldn't be terribly difficult.
I could go in and do this manually. But, again, let's take advantage of the AI where we can. I'll just hit refresh. We'll see what it's coming up with. Okay.
So it's got the first one. Second one is good. Is it done with the episodes people for Ed Miner Matt Miner to the workshop episode? Added all the people to the episodes. Alright.
Right. So Matt Miner, Brian Gillespie, there's the workshop. All I need are the images, and we're not gonna generate those via via this workflow. Now what I also am going to do here, right, now normally what I would have to do is go through each individual episode. So if I look for show contains MCP, that's gonna be case specific.
So I would have to go in and generate the transcripts for these, but can you generate transcripts for all those episodes using flows? Let's see what this comes up with. And, again, look, I've this is my first time doing in this. I've not practiced this at all. We'll see what's gonna happen.
There's a generate transcript flow that's designed to fetch Vimeo files and send them to Deepgram for transcription. This is exactly what we need. So it's gonna go through and call these flows. So now if I, like, go through and I look at the generate transcript, I can see that this has been triggered for that, for that episode. It's gonna continue with the rest of the episodes, and, that's a wrap.
Right? So what I'm able to do here is just take a a literal ton of busy work out of the equation for me. And now all I gotta do is hop in Figma, design some images for this, and I could publish seven videos in record time. So that's it for this video. I guess this will be the eighth video that we upload to this MCP.
But I can't wait to see what you guys build with this. Please share your feedback. Send us questions, comments, ideas that you have for, Directus and AI. I'll catch you on the next one. See you.