I Tried a Multi-Agent AI Developer Platform | MetaGPT x (MGX)

Jon Law
434 followers
Try MGX now:
https://mgx.dev/?utm_source=influencer&utm_medium=jonlawedu&utm_content=youtube
--------------------
Clips channel @jonlawclips
I do NOT sell any products of my own. Watch out for scams.
Queries: [email protected]
@metagptx
#MetaGPT #DeepWisdom #MGX
#MultiAgents #Software Dev #Developer
You know, I feel like we've been talking about AI a lot on this channel recently,
but I have no regrets. This is another AI video for y'all.
So I know multi-agent AI development platform is a lot of big words,
but this platform is pretty awesome and I'm really excited to show it to you.
So first, I'm going to tell you what it is, then I'm going to tell you who it's for,
then we're going to go build some stuff, and then somewhere in this video,
I'm going to sprinkle a little reward in for people who
watched all the way through just as a reward from me to you.
And it's not going to be in the chapter description anywhere,
so you have to actually watch all the way through to get that reward,
which may or may not involve some money.
And that said, cheers to you, wherever you're watching this from,
and let's get into the video.
So MGX is a multi-agent platform that's built on real-world software SOPs.
And I think that right there hits, first off, a really important point,
which is multiple agents working together that are each specialized to complete a larger task.
All that said, a lot of the chatbot AI tools today fail to hit some of those marks that are
required to translate AI from a personal tool into something an enterprise or a business can
actually use.
So what this platform does differently is it has a super-based integration,
Git versioning, full-on transparency, and then archiving.
So those are all things that are much better for an enterprise or a business or a coding team
that needs to work together to get real-world things done.
So this is meant for anyone who works in coding.
If you don't work in this space of software, this is that entry level to helping you do so,
but more so if you already do, whether you're a small business or an enterprise.
It's those two things I talked about.
So the agentic specialization and models working together,
and then the real-world software SOPs, all of that merged into one platform.
That is the angle that MGX brings to this game.
That said, we're going to go dive in the platform so you can see if it would be helpful for you or
not.
So I have three demos to share with you.
Let's go jump into my computer to get into MGX.
Okay, here's the MGX website, and I want to introduce three demos that we're going to
take a look at.
So number one here is the Link Hub, and I'm going to click on Link Hub.
So the agent building this is named Alex.
Now you can note the terminal right here, app viewer, editor, and then files.
So you can see this little link hub that was made.
I'm going to tell it to customize that to me.
So now I've given the agent some information about me and also uploaded a photo, and now
here in the actual files, we can see all of that being updated so that the link hub is
personalized to me.
So there we go.
You can see my profile image in there, name, location, email, and then all the social links.
If I click right here, then we can see the new tab.
Now this was just the appetizer just to warm up because what you see right here is that
we're using one agent to do this, the engineer.
If I click mode switch, I can go to team, and that is the bread and butter of NGX.
And that's what we're going to look at in demo number two.
So what I'm going to do, I'm going to say, hey, Mike, Mike is the team leader.
I want a website built for John Law.
You can Google him to learn about him.
He's a writer and creator.
So I'm just going to leave it like that as a very open-ended prompt.
And then this setting right here, this little mode, we used to have it on engineer.
Now we're going to have it on team.
So I'm going to keep the model the same, and let's get this started.
So here we can see Mike as the team leader, and Mike is now thinking.
And you can see right here that he has assigned one agent called David to first research me
in this case, and then Alex, another agent, will build the website.
So we can see Emma, a product manager, Bob, the architect, Alex, the engineer.
These are different individual AI agents that are all specialized in something.
And here we can see a little demo of me that one of the agents went and found.
So this is my background.
And oh, they have that I lived in California and in Boston.
So that is fairly accurate.
And it's looking at some other people as well.
And now it's going to go ahead.
David's being pulled in for data analyst work.
And it will pull this together into the final project.
So now what's been happened is Alex, the engineer, has been pulled in.
So the research report was saved in the code.
And then the other agent that's coming in to build the website is going to actually
reference that research report so it knows what to put in the website.
And we can see that the website will be built with this stack.
And in this case, probably Emma, the product manager, and Bob, the architect, won't be
pulled in for this kind of website build.
And we can start to see the website being put together here.
So it has my name, a little bio, explore my books, get in touch, some stats about me,
my age, where I'm currently located, education, key achievements, and the list goes on.
They even have, I actually, this is pretty interesting.
Yeah, all my books are here with the topics and a little description.
And now Mike, the team leader, has told me that the project has been completed.
And then what I can do is click publish to actually deploy this.
I'm going to open that link right here.
And this is the final website, along with a let's connect button, get in touch.
And even it looks like my actual YouTube is linked here.
And that was all from a single prompt where the agent went and did research about me.
Now I want to pull in Emma and Bob.
So I'm going to say, hey, Mike, can you come up with relevant tasks for Emma and Bob that
relates to this project?
Maybe something like a product unlock could sell and Emma could research that.
So it looks like the agent for Emma is going to go ahead and create or suggest a product.
And then Bob, the architect, is going to design the system architecture for that.
So let's see, Emma is going to create a product requirements document,
a PRD for potential digital products that I could sell to my audience.
And let's see how this goes.
OK, so that being said, Emma finished her report.
That's also stored in a file here.
And then now Bob, we can see right here, is going ahead and designing the architecture
for implementing the product that Emma suggested.
So it looks like one of the things that Bob put together was this little sequence diagram
for the whole project.
We can see the user all the way through front end, API gateway, off user, course, payment,
video.
And it looks like this is, of course, is the suggested product, which makes sense for sure.
And then under the research file, we can see the research report that was generated.
So biography.
And then we can scroll all the way down.
This is all about me.
And then here is the product requirements document that was generated as well.
So it's a financial education platform.
Totally makes sense given the audience product definition.
And I can scroll through here and you can see all kinds of things.
And oh, that's interesting.
We even got a competitive quadrant chart.
We have the different tiers, revenue projections, and go to market as well.
So that right there is a little bit about how to use the team effectively to go ahead
and design projects from start to finish and how to use these models together by prompting
the team leader.
Now, if you think any of the features that you have seen thus far can be a good fit for
your business or for your personal passion or hobby, and I would bet that that's the
case, then you can go experience the agent team within MGX today.
To do that, just hit the link in the description or visit deepwisdom.ai to get started.
That's the whole multi-agent workflow with the super base integration, get versioning
and comprehensive archiving that you get along with everything else that you've seen today.
Now, as a fun third demo here, what I want to do is ask Mike, come up with a new SaaS
product idea.
I want you to utilize all the agents on the team and come up with something specific.
And for a very targeted B2B audience, we can set up payments and off and focus on driving
value through the software.
So I'm going to keep that on team and then let's go ahead and see what kind of business
it comes up with.
And then as a follow-up for this video, if you guys want to see it, we can continue to
work on this and then actually run ads as an experiment to the product with Stripe integrated
to go ahead and see if we can use AI like so to actually start a real business.
Let's go.
So let's just let it run for a few minutes and we'll see what the product is.
Here we go.
So the platform is quality management and compliance automation platform for SMEs.
67% of SMNs use spreadsheets for quality tracking.
And I believe the recommended price point is $49 per user or per month.
Okay.
The team's been working for half an hour or so, and it looks like it's ready for launch.
Now, what I'm going to do is test the app.
It looks like this is the homepage and it gave us a demo account right here.
And this is all with one single prompt.
There was no follow-up build there.
Let's go and see if this works.
So the sign-in works and cool.
This is what the dashboard looks like.
Compliance status, whatever CAPAs is, quality score, audits, recent activities.
So all this data seems like it's in here just so you can see for the demo account what it
looks like.
Quality metrics, different things we can do.
So start a quality inspection, report a quality issue, compliance.
And let's go Google what CAPA is quickly.
A corrective and preventative action.
Okay.
And this is the demo business.
We have a compliance page.
We have quality.
And then we have audits.
Looks like that hasn't been built yet.
And reports, which is also yet to be built.
And let's see, we can go to profile.
We have billing.
Oh, it looks like billing is actually set up.
So we're currently in the professional plan.
$79 a month and there's enterprise and starter.
And then let's go to settings.
Well, it looks like that also hasn't been built.
Cool.
So that's a great demo right there.
And MVP, what I'd say is to the bot, hey, Mike, looks good.
Let's go ahead and finish up all the pages that say coming soon.
So it's going to go and build out the admin dashboard supplier management pages.
And it looks like a whole list of features here.
Now, if you want to see this turned into an actual app with real value,
solely through this agentic team, let me know.
And we can go ahead and do that in a future video.
So again, go visit that link in the description or the one on the screen.
Now that said, that's all for me.
I'm going to finish off this monk fruit Red Bull.
And I'll see y'all in the next one.
Cheers.
