Headless LMS Ops: Your Academy with the Skilljar MCP

All right, thanks Dean. I think I only give him half that information. So I think half of that he made up or AI did. I'll figure out which portions.

Anyways, I'm Nate Whitley, I lead the Skildra product team, excited to meet with you all today. I'm going to be talking about Headless LMS Ops. I am going to do a few slides. There's not a lot here and then I'm actually going to get into a demo.

So most of the session is going to be me demoing how to use our MCP server and some of the use cases around that. So today is going to be a little scary on my end, as long as Wi-Fi holds up will be good. All right. So as I mentioned, I'm going to be doing a little bit of introductory here.

I thought about doing a poll at the beginning here, but I didn't think what it really comes down to is when I'm meeting with customers, more than half of them are still asking, like what is MCP? What does this mean? How does this, what does this mean for me? And so I do want to start off with just like what is LMS, Headless LMS Ops?

What does this term mean? What is MCP? I want to get into some sample use cases, to give you a sense of how this can be used, and then I'll get into the demo itself. So only about 10 minutes of slides, maybe less, and then we'll get into the demo.

And I would encourage everyone, if you're kind of in the back, you'll probably want to sneak forward when you get into the demo, just so you can actually read what's going on here. All right. So Headless LMS Ops. Really, when you think about this, there's a left-hand side and the right-hand side.

There's a today and a tomorrow view of where things are at. The today is really the experience that you've all had with LMS. LMS is like a skill drawer over the last 10 to 15 years. It's ultimately admins log into the dashboard, and that's where you do your day-to-day tasks.

You manage enrollments, you publish content, you pull reports. When you think about tomorrow, you're ultimately getting to a point where you can operate skill drawer with your AI tool of choice without ever logging in. You do not have to log into the dashboard in the future. You use what tools you want, you connect with our MCP Server APIs, and that's how you operate on a day-to-day basis.

When you think about what's going on within the dashboard, you're doing one task at a time. You're going page by page by page, or you're going down a list of things that you're trying to accomplish. With your AI tool or any MCP-compatible agent, you can run your full LMS operation natively. What that means is you can start to spin up repeatable workflows that you can operate without you having to individually go through the list one by one by one.

Then when we start thinking about analytics here, in the today space, you surface information manually. You're going into the dashboard, you're pulling reports, you may have a data connector so you have a BI environment, but you're ultimately going somewhere trying to find information, you're trying to find patterns that are important for you. When you think about what the future looks like, AI reasons over your data. You hook up your AI tool of choice into your BI environment, or you're using MCP calls from Skill Drawer to understand who's completing this content.

What does that pattern look like? Does that benefit us? Is it resulting in higher retention? AI can start processing this information and then recommending actions and not just recommending them, but it can actually execute on those actions for you.

What is MCP? The words literally mean model context protocol. That's what MCP stands for. The model is AI.

This is Claude, ChatGPT, Gemini. There's many others. The protocol is going to be the standard. The example that I like to use is USB.

Before USB-C came around and standardized things, you had Lightning cables, USB micro, USB A, you had all these different cables and there's no real standard for what you needed to charge your phone or charge your laptop. The protocol here is basically that standard. It's providing how machines talk to each other through APIs. Then the last is going to be context.

Context here is a key word. This is ultimately the information that you give AI so that it can actually understand what you're trying to accomplish. This is going to be your data, your tools, your systems, your academy, and the like. Let's get into sample use cases.

I'm in products. I just wanted to give you an example of things that I do on a week-to-week or daily basis. These may not apply directly to you, but they should give you a sense of how I am using these types of tools every week. When you think about research, a key part of what I do in product is I'm doing discovery.

I'm trying to understand customer problems, opportunities, and where those opportunities exist so that we can prioritize them as a part of our roadmap. There are many different ways that we collect customer signals in the world of product. The example that I have up here is we have Gong transcripts. We have Gong recordings of our calls.

We have Jira tickets that are submitted from our team members. These are ultimately become direct feedback from customers. I use Cloud Code to ultimately understand what's going on here. I can just natural language query these to understand where some of the potential opportunities are.

Another one is going to be win-loss insights from deals. We have a tool that we just got MCP access to. Now I can just ask it, pull down like what's happened in the last month, tell me where the opportunities are, are there any trends that are happening like that? Then the last one in here is I fucked up Cloud into our data warehouse.

I can now understand what features are being used in a cleaner way with our customers and associate it back to the churn or retention signals. I can also better understand usage patterns that exist for features that are being used across our customer set. Again, this is all happening through my AI tool choice which is Cloud and using different MCP servers across tools. One that might be more applicable for everyone in this group is going to be meetings or meeting prep, understanding what's going on with your customers or even internal meetings.

An example here is going to be around pulling customer details from RCS or Staircase MCP. I use this when I'm prepping for meetings to meet with customers like you, trying to understand what is this customer asking about, where they add, what's their sentiment, what's their health score. Another example here is we also have Notion again. So it helps record some internal calls.

I've hooked up the MCP to Notion, and I can pull down meeting notes. I can ask it, what are my action items for the day or for the week to make sure that I'm staying on top of things. I'm not logging into any of these tools here. This is all just happening through my AI tool of choice.

Then on the far right, just to give you a few more examples here. So we use GitHub and that's ultimately where our code repository is. What we've done now is we've hooked up Cloud into our GitHub repo and Jira tickets. As the engineering team is completing their work, I can now go ask it to build release notes off of the changes to the code.

When I say I, it's my team. Our team is now doing this to generate release notes once code has been committed without us having to go think through and build out the pretty lengthy documentation that exists for that may be on our support articles or that we're communicating to go to market teams. Then the last one here is actually hooking it up to Slack. So when I'm doing some of the research over on the left-hand side, sometimes I'll just say send it to Yoris.

All of a sudden, it just sends this chunk of research that I just completed off straight to Slack. I'm not even having to directly draft up an e-mail or a note to him. Same thing, it can actually happen the opposite way. If Yoris were to send me a set of research, I can have it go pull down the research into Cloud, process it, let me do my thing, and then I can actually move on from there.

So I'm not having to copy-paste things across different systems, I'm just directly hooking up through MCP. So those are some of my day-to-day weekly use cases. Let's talk about some Skill Drawer MCP use cases. These are going to be some of the things that I'll be demoing in just a few minutes here.

So the first of these is going to be content creation. We have API endpoints that allow you to create courses, to create lessons, create quizzes. These API endpoints are going to be tools that exist within the MCP server. I'll be able to directly show you in a second, I'll give it some source content, and it'll then create the content locally within my Cloud instance, and then it'll actually push it to Skill Drawer.

I'll try to do a side-by-side here during the demo, so you'll see Cloud doing its thing, and then you'll see it says, "Create course." I can go into my Academy and I'll show you, course has been created. You can see it as it's iterating through this, lessons being created and pushed back out to Skill Drawer. Next one is going to be Enroll learners. This is actually a little bit of a combination with the next one, but we also have API endpoints that will exist as tools within our MCP server.

This will allow you to enroll learners into content. I'm going to actually combine it with the next one, which is Manage Groups. You'll see me actually go find users in a group, and then I'm going to enroll them into a content that I just created. This will allow you to see how you can combine the tools that exist within the MCP server to do some of these student management related use cases for Skill Drawer.

Then the last one is going to be Translate course. This is going to operate similarly to content creation, but Translate course here is going to go read a course content, pull it down, you select the language that you want it to do, and then it will recreate the course in that language in Skill Drawer. So this is all happening in the AI tool, leveraging the MCP server that exists. All right.

There's one more concept I'm going to introduce, and this is something called Skills. So a skill is a reusable prompted workflow that runs inside Claude invoked with a skill name. This is also something that OpenAI is adopting, and so this is becoming a standard in the AI native labs. So really what this gets down to is a skill and codes, a specific multi-step process.

So what data to gather, how to reason through it, and what to produce. So you can think of this as a standard operating procedure, or almost like an agent if you will, that allows Claude to follow on demand that provides consistent repeatable and tailored to your tools and context. The key thing with this skill is, so the example here is going to be course builder. I built a course builder skill.

Like you don't want everyone on your team to have their own distinct way of that they build courses. You want a standardized way for them to follow this process. So the course builder skill that I built, that I'll be demoing in a second, there's a core skill itself. This is where the skill definition lives.

So I'm using the Addy framework for instructional design. So that's baked into it. It's going to have tool calls that knows what tool calls to use, like when to create a lesson, when to create a course. There's gates in there.

So an important part of this is going to be a human in the loop aspect to this. An important part of using these types of tools is verify, don't just blindly trust. So human in the loop gives you an opportunity to provide certain tool gates to block or stop the AI from continuing on. So you'll see a couple examples of this in the skill itself.

I'm going to have it show me a course outline before I have it actually go build a course, for example. Then there's two key reference files that I've associated to this. This is my own style of how I wanted to build the skill. But the two reference files here are, there's a course guidelines, Markdown file, MD is Markdown.

So think of this as your configuration file. This is where you can set design or brand guidelines. This is where you can set reading level. This is where you can set knowledge check patterns.

This is where you can set things like, things you don't want to include like mDashes. You can include brand tone, brand voice. So this is basically your way to customize the skill to fit your company. The skill can't just be like drag and drop straight into your company.

If you were to use this, you'd want to customize those things specific to your company. Then the last reference file here is HTML patterns. This is basically the skill that I built here is going to be creating rich HTML lessons and these HTML patterns are basically predefined custom CSS like tabs, accordions, hover over effects, specifies table structure, table formatting. So basically is giving it all the guidelines on how to structure the lesson itself.

So these are both configurable reference files here, but these basically give the skill, the understanding on how to create the content, what that content needs to look like. So a skill and I think this is a really good example here. This is a more complex workflow because it's going to be multiple steps. It's going to take five to 10 minutes to execute.

Again, it's really around having that standard operating procedure that can be repeatable and shared across your team. Last slide before the demo, I just want to cover a few things. So I'm going to be using Cloud or Cloud Code rather. I'm going to be using the desktop app.

You could do this in OpenAI or Gemini in the future. So that's what I'm going to be using here. I'm going to be doing a side-by-side. So you're going to see or it's a little bit of an overlay, but you'll see Cloud running in one side and you'll see my Academy in the other.

So as it's starting to do, actually push things in a skill drawer, I'll do my best to move around a little bit. There's going to be some context switching here, so you're going to have to stay with me. It's more just to show like magic is not just happening. There's actually tangible proof that this is running.

I'm going to, there's actually two skills that are going to be running here. There's a course builder skill which I just talked about a second ago, and there's also a translate course skill. The translates course skill here is basically going to follow the course builder structure, but it also is pre-loaded with the language packs that skill drawer supports, so it knows which languages to do, for example. Then I talked about this on a slide a second ago, but I'm going to be creating a course from source content.

The skill itself is built in a way where it could be a Google Doc, a Microsoft Doc, a PDF. I'm just going to give it a support article URL. I'm simple in that way, so I'm just going to give it a support article, and that's going to be the source content that it's going to build off of. I will also publish a course through the MCP.

Generally, I think the workflow would be you're going to create the course, you're going to go into the UI and review it before we publish, but I'm just showing you you can publish it, enrolling learners into that course, and then translating the course into a different language. So those are going to be the core workflows that I'm going to be demoing here in just a second. That's going to be the time where we can flip over to my home screen. In this case, I'm directly invoking the skill, so I'm doing a slash and the name of the skill.

It will work if I just say I want to build a course, you can use natural language. I'm just showing you what the skill structure is here. This part is going to be me talking while it's running for a little bit. So what's going to happen right now is it's going to go fetch the content.

So it's going to that website right now. This is a support article. Like I said, you could do it through G Drive or something like that. It's fetching the content.

Once it fetches the content, it's going to basically process or understand the content. From there, I have a gate built in. It's basically going to ask me three questions. The three questions are going to be, who's the audience?

Audience is stage. Review preference. This is whether or not I want a gate here. Do I want a final quiz?

I'm going to ignore screenshots, but I could have it pull in screenshots from this content too. Let's do new users of staircase. Is the audience? Review the course outline.

Again, this is just the skill. These are the things that I want to always have some say in as after it pulls in the information. It's going to take 30-ish seconds here and it's going to generate a course outline. So I'll show you what the course outline looks like just to give you a feel for how the skill is operating.

It's reading the reference files that I mentioned earlier, the course guidelines one. When I'm at home, it takes like 25 seconds, it's a little faster. The course outline is going to pop up here in just a second. Again, you can see right above this, I wanted to review the course outline.

If I was fully trusting, I could say it could one-shot it and just go straight through and build this. But I want to show you how human-in-the-loop interactions exist here. Here we go, the course outline. Here's the course title, here's the short description, and then the long description.

I'm forcing learning objectives to always be stated in a three to five bullet format for learning objectives. This gives me a chance to look at that and then it's going to give me five lessons. Sorry, I'm making this big font. Lesson one, introduction, MCP and initial setup.

It's saying content type is pros and info call out. So it's telling me the structure of the content, and then it's saying there's no knowledge check necessary for this one. So this is all clawed through the skill of making these determinations. You go down next lesson, lesson two, connecting chat GPT to staircase AI, content type.

It's going to be an order checklist with a progress bar. This is using the HTML patterns that I was referencing earlier, and then it is going to create a knowledge check for this. So it's basically making determinations. Is there enough content here?

Is it worthy of having a knowledge check or should we skip the knowledge check and go into the next lesson? All right. There's five lessons, and then there's going to be a final quiz which is technically a six lesson. The final quiz is going to be 10 questions, 80 percent.

So I'm just going to say go for it. This is now going to take seven to eight minutes probably, somewhere in that neighborhood for it to run. Let me just explain some of the things that are going to happen. I'm going to actually trigger the translate course on another course while we're waiting just to let it go in the background.

All right. So what's going to happen is it's going to generate my lesson content. Because there's more than three lessons, what I've done in the skill is I'm actually going to parallelize the work. So there's going to be sub-agents that are going to be building each lesson.

If there's a knowledge check, it'll also be responsible for the knowledge check of that content. Since there's five core lessons, once those five lessons are done, those five agents, another agent is going to come back through and do the final quiz on all the content above it. So that's the logic behind the skill. It's going to create all that locally, and then it's going to go do it, push it into Skill Jar.

So it's going to create all the content first, and then it's going to push it into Skill Jar. So it's going to take a couple of minutes here for it to do its thing. I'm going to go find course. So this time I'm not invoking the skill itself, I'm just natural language typing to it.

It's probably when I've done some testing, it's going to hit a couple of different tools until it finds the course. So I'll come back to this in a second. But this is going to go actually faster than the creating course, because it's going to read the content and then translating it. It doesn't have to think about creating content and the structure of the content.

So the course builder skill actually takes about twice as long as the translate skill here. So it's fine, it's looking for it right now. Come on. I'm pretty sure this is the one here.

All right. Well, it thinks there. So I'm back in the course builder skill now, so I just switched sessions here. So it's saying all five agents have run, all lesson content is ready, it's building the final quiz right now, it's actually showing me in the chat the 10 questions.

I think I've instructed it to only do multiple choice in true-false, so there's only two question types here. So it looks like this is question number eight. So it's basically putting it all back in here. Now it's going to start creating the course in the skill drawer.

So it's about to start invoking more skills. So yeah, creating your course in skill drawer. So you can follow through, it's basically telling you slowly what it's doing as it's executing through this. So creating course, I'm going to exit out of this one.

There's our course shell. This is what we looked at earlier. So this is the course shell. So it creates the course first, it brings in the course title, short description, long description.

No lessons yet. So we're going to see, I'm going to slowly refresh this occasionally, and you'll actually see as lessons are completing in my Cloud session here, you'll see the lessons start to show up. Again, everything has been already created locally in Cloud. So now it's just working through the API calls itself, using MCP to make those API calls.

So we'll see it here in a second. It's going to finish, it'll give me the lesson ID. Once I see that, I'll know that I can go look in the system here. It is not liking me right now on this one.

All right. I'm just making it do an extra step. In reality, what I would do is I would have given it in one of the configuration files my main domains, I would have made the list there, I kind of skipped that step. So it's like having trouble finding which domain I exist in right now.

All right. So it should be done with the lesson here pretty shortly. I'm just checking that out. All right.

So it found it now. So it basically had to make an extra tool call. It had to go look up my domain, look up all the domains that exist on my Academy, found the domain, then it went and looked up, use that domain ID to look up courses. So it had to take a couple extra steps there because I didn't give you enough context as ultimately what happened.

If I'd have provided that information, it would have gone a lot faster here. All right. So I'm in the translation session right now. I should have named these over here.

So it just pulled all the content, it fetched all the content, and now it's translating it into French. So it's doing it in a very similar way and pattern that exists. The skill is also instructed to follow the same structure, the same number of lessons. So it's basically trying to mimic the exact look and feel, it's just changing the translation of the text.

So this is going to win by a couple of minutes probably once we get there. Let me zoom out quick. All right. Here's the French course.

So this is again just the French shell, and then it's going to start executing on the lessons themselves. In real life, I would not be babysitting this and watching it. I would just come back after 10-15 minutes and just see that it's done. So just realize I would run this, set it in the background, go do my next task, and come back to it a little bit later.

So this is not really a true workflow per se. It cannot right now, no. But the HTML patterns that reference file that I mentioned earlier, it's mimicking the types of things that can be done in SCORM. All right.

Let me go in here. All right. So we got some lessons. So this is the main course that I was building.

This is the course builder. All right. So let me pull up. So I'm in preview mode.

This is more what it's intended to look like as an end-user. If I stay in the Admin dashboard, it still has a little bit of differences here. All right. So this is the first lesson.

So there's some boxes in here. I didn't really pay too much attention to what I was trying to do. All right. So here's a checklist example.

Again, this is all just rich HTML. So it's following a checklist pattern. You can see as I was clicking through these, it was marking off progress here. If I unclick one, it unclicks it over here.

So there's a little bit more interactivity that exists with this. There's a knowledge check down here. There's three questions. Well, what do you know?

It happens like one out of 10 times. All right. So there's an all check. Here is more of a tab-based lesson type that it's creating.

So there's content underneath the tab. I'm in a weird format here, so I can't really see this too well without scrolling a bunch. So just trying to give you a feel for the types of things that can be done with this. Again, largely it comes down to the configurations that you and your team would want to do and incorporate into this that meet your standards.

So there may be types of HTML patterns that do or don't make sense for your team, and you could just make tweaks to that or add onto it based on what you're looking for. Here's another checklist type example here. So I'm going to go back in here. I like this course.

So I'm going to publish this now. Which one is this? That's French. Let me show you the output here.

So this is just finished. Again, this is the course builder session. What it does is it's telling me here's the course, this is the title, here's the ID, which doesn't really make it too important. This is the exact link that would take you to the dashboard of that course in your dashboard.

So clicking here, you don't have to be in your dashboard the whole time. You can just use this link, it'll take you straight into that course that was just created. And then here's a quick summary of what was just created. So it's saying, here's lesson one, it's just HTML.

Here's lesson two, it's modular, meaning it's one of the stacked lesson types, which is HTML and a knowledge check. And so it's giving you a quick rundown, and then the next steps are you go review this, click on the link, review this in the dashboard. You can make edits there if you want, or you can directly publish from there. I'm going to skip that step, and I'm just going to publish from here.

It's just going to follow default publishing settings. I'm not trying to put it on any special pages here. It's probably going to have to go find my domain. Okay, it looks like it's doing that right now.

And so it's going to publish it, I'll show you in a sec, it'll show up here. Once that happens, I'm going to start going through and find users in a group and then I'm going to enroll them in this course. Again, I didn't give it the domain specificity here, so it's like working through the logic on its own without me having to tell it a bunch of information. All right, you can see it just got published.

Here it is. All right, let's see. Actually, let me do this. I'm going to chain a few things here.

All right, so earlier I had enrolled and then completed 20 learners into this course. So I'm actually just going to go ask it, tell me who completed this course. I think it was 20. So it's going to go hunt for this.

Based on this, I'm going to add them into a group, and then I'm going to enroll them. So I'm going to chain a couple of different things here for you all. So this is actually, you can see these are all the different tool calls that are happening behind the scenes. So I'm opening up, you can see the list here.

So even though it doesn't look like that long or that hard of a prompt here, it's just six or seven words. It's actually three or four different steps that are happening behind the scenes to execute this. It doesn't quite warrant in my mind having a skill around this, but it's something that, again, just natural language is going to execute on multiple steps for you. This one might take longer than I thought.

All right, there we go. So we got 20 completions, which is what I mentioned earlier. These are all test users. I've created dummy data in here.

All right. All right, so I want to create a new group called Staircase, and then I'm going to move these learners into that group. So this is now going to go execute on a few different calls again. So this is showing you the types of things that can be done with the MCP server.

Once this is done, then I'm going to actually enroll them into content. All right, just created the group. You can see it's pretty quick when it's doing some of these simpler tasks here. All right, all 20 of those users were added to that group.

All right, and this is going to be a little bit interactive here too. So you can see there's no enrollments, there's no activity in this course. This is the course we just created, just published, so nothing's happened here. All right, so again, those 20 users that I just moved into that group are now all going to be enrolled into this course.

So you can start to see how you can start to combine certain workflows or look for certain patterns and behaviors, and then move students around based on what you're ultimately trying to orchestrate for them. And then I'm doing this prompt by prompt by prompt, but you could create workflows where you're running this on a nightly basis and you're looking for certain patterns and based on behavior or attributes. And then you're moving the users, these students, into those groups or you're moving them into different enrollments based on those patterns that you're detecting. So this starts to combine some of the concepts that I talked about on that headless LMS ops slide of the future where AI can start to look through your data and understand what's going on and start to act on that data based on recommendations or guidance that you give it.

All right, yeah, we're already at 11, so you can see it's now just running one by one by one by one as it's enrolling all these users. And then once it's done, it's gonna give me the list here. You can actually see it gave me a little bit of a summary here too, where it's basically telling me all 20 got enrolled, there are no failures, no duplicates. Some of the logic here is making sure that we're not duplicating enrollments as we're executing on this, so it's actually knowledgeable enough now to not do that.

It'll actually skip a user. And so it would have said 19, but one would have been a duplicate, and so that's why it skipped it. So it actually has, because we're using these types of AI tools, it has that kind of intelligence behind it to understand certain logic like that. All right, this should be done, we should be at 20.

All right, and then just to kind of show one last thing here. All right, and then the last thing that I'm doing here is I'm just gonna complete those 20 enrollments, all right? So this might be manual processes that you would do on a typical basis, but I'm at least gonna show you that this is one of the last steps that you can do here if you wanted to issue certificates to a group, or they attended a live event and you wanted to follow up on that. All right, this is gonna be the last thing that I will be showing here, so we can open it up for Q&A in just a few minutes, in a minute here.

All right, you can see we're already at 16 of 20 completed, and we're done. So here, I can pull that back up. All right, so I went pretty quick. I was context switching, showing different windows or sessions, but I wanted to give you some sense of some of the use cases that are gonna be unlocked with this type of capability.

And so I'm happy to open it up for questions if you have any answer what I can. You heard a little bit of a teaser this morning on this being available. You'll hear more about it in the keynote tomorrow morning. But we are going to an open beta on our MCP server, and we are gonna continue to invest and iterate on this.

So some of the capabilities like quizzes are not quite there yet on our MCP, but they will be there here in the near term and will be continuing to add more tools and API endpoints for those capabilities. Is that beta version of it? So it's actually a local MCP that I built myself. When I say I built, I'm not a coder, but I use Cloud to help me iterate on it.

It's largely because we've been working up to right until Pulse to have the beta ready. So it's intended to mimic what we want the beta to be in a week or two. Yep. All right, thanks, Nate.

We have a lot of questions that have come in here in the last couple of minutes. We probably won't be able to get to all of them here, but I will go from top to bottom here. So first question, are there any companies using Skill Jar completely headless now, so via the MCP or API? And what are some cool things that they're doing?

There's one in particular that basically logs into the dashboard as a last resort. They do not log into the dashboard. And if I get a feature request, it's an API. That's what I get as a feature request from this group.

And so they are creating content in this way already. They're just not using an MCP server. They're creating skills to wrap around the APIs and they're just kind of doing that on their own. One thing that's interesting for them that they just started doing that I didn't demo here is password resets.

And so they actually just created a skill with one of our new API endpoints on password reset, so they get support tickets in. They don't have to go into the dashboard and figure out where to go and click this. They just have a skill that runs. They just drop in an email and it sends a password reset email to the users.

So when you think about the set scale, when you have tens of thousands of users on a monthly basis, it starts to add up when you get maybe 100 or 200 of these every month, and so they are just looking for ways to reduce manual time spent on the redundant things. I hope that gives an example. But course creation, they're already doing. They're doing some of the managed groups.

They have jobs running in the middle of the night, basically, that are finding people that did something and then they're moving them into a group and enrolling them in content. So they're chaining this, chaining this logic together in some of their workflows. Very good. If you're a content creation tool, such as Aparta I.O., how would your application change?

How would your application change if you were- Oh, sorry. Reading's important. How would your approach change? How would your what, sorry?

Approach. Approach. Approach. There's still going to be a place for those types of tools.

I think there's also an opportunity for us to enable each other to use our MCP tools themselves, like being able to translate content, for example. If a tool like Partha did not have access to that content, you could use Skillshare to go read the content and then create or translate it into Partha as a SCORM package. I think there's also going to be, those tools specialize in content creation. And so they are going to have more interactive elements that exist.

This is just providing a different way or a new way for teams to explore creation of content that wasn't natively possible in Skilljar previously. Very good. Are these skills you shared included in the Skilljar MCP release or do we need to go get them separately? I feel like you put that question in there.

Yeah. The skills are not available right now, but my intent with these skills is that we can make them available to our customers here in the near future. We have to figure out the right way to make them available. Are they downloadable?

If we make updates, what does that mean for those skills? So there's things that we have to kind of solve for in making these extensible. I do think the two in particular that I showed are really key skills that would make it easier for customers to get started. If you just use the MCP and you just tried to create content, it'd probably be kind of flat content.

And so having a skill like what I was showing is going to make your content more interactive, more meaningful, better fit your brand guidelines. And so I think these skills are going to be really important for the successful usage of our MCP capabilities. So my intent is that we'll be making these available here in the near future. Very good.

Enrolling people in a course from Claude is a cool capability. Why would I do that instead of just doing it in the skill jar platform directly? I think it comes down to bulk. And if you're doing repeatable, like if you're doing this on a repeatable process where you're doing manual enrollments weekly, daily, I think this is really about removing the redundancy and like finding efficiencies for your team so you can go focus on higher value things.

And so I really think about this in terms of like, if you have a list of 100 people that need to be enrolled, you just provide a CSV file in the cloud and it does its thing for you. So I think it really comes down to reducing the manual work. Now if you're only doing one enrollment like every month, like you probably don't necessarily need this. But if you're doing it on a repeatable way, this is going to make sense for you.

Very good. Does the future of learning include learner created courses? Think of a learner logging in and explaining what they want to learn. And for what content, this gives the learner literally what they want and feel they need.

So this gets into a different topic than what we covered today. I think this gets into something that we call learning everywhere as well as, it really comes down to, I view this as micro learning. And then using micro learning, the blocks of micro learning to potentially dynamically basically Lego block style build content that's specific to the end user. And so I think that's not really what's being solved here.

What we talked about here is all admin based use cases, like you as end users of Skill Drawer itself, not end users taking learning on a Skill Drawer Academy. So these are two separate things. I wouldn't bring these two things together. It's another topic, I'm happy to talk about it another time, but I'd like to see if there's another question around MCP first.

All right, now that sounds good. I think we have time for maybe one or two more here. You mentioned in the demo that something didn't warrant making a skill four. How do you make that assessment?

It's experimentation trial and error to a degree. I think you can kind of see what I was doing there. I was just asking it to do something and seeing if it did. If it's unsuccessful or it doesn't produce the right output that I'm expecting, then that might warrant me giving it more information, more context, so it can be successful.

And then it starts to lead into, is this workflow complex enough, or am I going to be repeating this workflow enough where I want to have kind of structure around it? And then that kind of points me towards, I want a skill for this type of capability. All right. It really comes down to context and trial and error and just seeing if it works.

All right, last question, Nate. It's a toughie and I think it's the most important one here. I believe the Skill Jar MCP is an open beta. How do we get it?

Yeah, there are some release and comms done earlier by our great marketing team. All right. And so there should be support documentation, a release documentation that was shared out earlier today that you can access. This is in a beta.

We are going to continue to add on to it. That also means supporting more and more tools where MCP Server can work with, different ways that you can set this up. As a final note, more likely than not, you will need to interact with your security teams at your company before you can get this set up. So this also can be a starting point for you to get going and get that process moving with your security teams so that you can ultimately use this here in the near future.

Weeks, not, weeks, not quarters, yeah. You heard it here, weeks, not quarters. Hold them to it. All right, everyone.

Let's give a big round of applause for Nate Whitley here.