The rise of the agentic economy on the shoulders of MCP

00:00:00.000 | So let me start with a question: how does intelligence emerge in biological systems, right?

00:00:20.240 | Well, it's through neurons, right? When neurons are born, they are just like individual cells,

00:00:26.120 | but over time, they grow their axons and dendrites and establish connections with other cells or other neurons

00:00:33.840 | and actually learn how to communicate in order to pursue their own interests, basically, like to get nutrients and so on.

00:00:40.200 | And over time, they learn how to communicate with each other and with other cells to get nutrients and basically thrive, right?

00:00:46.360 | And this collective behavior, if you zoom out and look at a large number of them, it's something we call intelligence, right?

00:00:54.320 | So it's like emergent behavior of smaller individual units that pursue their own interest.

00:00:59.880 | So how does intelligence emerge in the markets, right?

00:01:04.720 | People always talk about markets like, well, market things that, market reacted to this and so on.

00:01:10.880 | And in some way, markets are more intelligent than individual participants of the market, right?

00:01:18.640 | And it's their mutual interaction of these individual members of the market, basically, who pursue their own interest,

00:01:26.200 | and communicate and establish new interactions with others, where some sort of collective intelligence,

00:01:32.200 | which is bigger than the sum of different parts, emerges, right?

00:01:36.200 | So, how does intelligence emerge in companies?

00:01:39.760 | Well, this one is provocative.

00:01:41.760 | Through Slack, right?

00:01:43.760 | Where people interact and pursue their own interest in the company, and all together, the company, well, sometimes,

00:01:51.760 | becomes more intelligent than the individual employees of the company.

00:01:54.760 | And so this leads to my final question.

00:01:58.320 | So how does, or how will, the general intelligence emerge in computing systems, right?

00:02:03.200 | And there is a lot of talk about AGI and, like, you know, ever larger models exhibiting, like, super intelligent behavior.

00:02:10.720 | But in my opinion, the general intelligence will actually emerge through interaction of multiple entities, you can call them agents, basically like multiple models pursuing their own goals, interacting with each other, and all together exhibiting something which we can call general intelligence.

00:02:29.160 | And thanks to MCP, we finally have this missing part that allows the agents to communicate with each other, and really, like, create a fabric or agentic mesh where they can talk together.

00:02:43.160 | So, hello, everyone. My name is Jan Czern. I'm the founder of Appify, and I'm going to talk about the race of the agentic economy on the shoulders of MCP.

00:02:52.560 | Basically, basically, economy where agents can, you know, find counterparts to interact with and purchase services from other, from businesses or tools or other agents, right?

00:03:05.560 | So, like, B2A and B2B, sorry, and A2A.

00:03:08.760 | All right. So, before I start, let me just introduce quickly Appify.

00:03:14.960 | Appify is a marketplace of 5,000 tools called Actors, and historically, we come from the web scraping industry.

00:03:22.520 | Right. So, most of these actors are data extraction tools that allow you, you know, to get data from social media, from search engines, data for AI, for building rack pipelines, you know, data from web for lead generation, and so on.

00:03:36.120 | But also, there are other tools, like data processing tools, and so on.

00:03:38.920 | So, all together, there's about 5,000 of them, and some of them are built by Appify, some are built by our community of creators who actually make money on it, right?

00:03:48.520 | So, it's like a marketplace of software creators, if you will, right?

00:03:52.720 | So, Actors are self-contained pieces of software based on Docker with well-defined input and output, right?

00:04:00.120 | And basically, they represent a new way how to ship software and publish it, you know, and integrate to, you know, other systems, right?

00:04:07.120 | So, for example, Google Maps Craper, it's a quite popular actor from our store.

00:04:11.320 | It can extract data from Google Maps, right?

00:04:13.320 | More data than the Google Places API provides, right?

00:04:16.320 | Well, there is, like, creator of the actor description, you know, different stats and so on, something you would expect from a normal marketplace.

00:04:24.320 | And actually, thanks to the way Actors are built, it's actually super easy to integrate Actors from other systems, right?

00:04:33.520 | So, for example, we have SDKs for TypeScript, for Python, for OpenAPI, for CLI, like we can call it from Terminal, and it's only because they are well-defined units of software with input and output, right?

00:04:47.520 | Also, we have integrations with workflow automation tools like Manic, Zapier, you know, Clay, and many others, so to make it really easy to call Actors from these systems, right?

00:04:59.720 | But obviously, now, we also have MCP integration, which makes it possible to call Actors from AI agents or, you know, AI workflows.

00:05:08.720 | And the way it works, actually, is the agent just needs an API key or, you know, a workflow account on Appify, and then through our MCP server, basically, it can interact or call any of those 5,000 Actors on our Marketplace, right?

00:05:27.920 | And actually, this only became possible thanks to, I would say, the killer feature of MCP, which is the tool discovery, right?

00:05:35.920 | Actually, not many clients support this yet, but just yesterday, I saw that VS Code added support for it, and actually, just like two days ago, Cloud for Desktop added support for tool discovery.

00:05:49.120 | And basically, how it works is that the client connects to the MCP server and dynamically discovers tools to use and to interact with based on the workflow, right?

00:06:03.320 | And let's say we have, like, 5,000 tools on our store, and there is simply no way we could publish all these tools through OpenAPI, because, you know, the context would be just too large, and, like, the more tools you have, the, you know, riskier the result is, right?

00:06:16.520 | So we really want, like, provide the tools only, like, as needed, and that's only possible through tool discovery, which I think is really the main thing that will actually make MCP really a huge differentiator from OpenAPI, for example, right?

00:06:34.720 | So, MCP actually quickly became a standard for agenting interaction.

00:06:39.520 | This is Google Trends data showing that MCP is basically dominating the space compared to OpenAPI or A2A from Google, right?

00:06:47.920 | And actually, I think MCP already became a standard for agenting interaction.

00:06:52.920 | And it became so popular that currently there are, like, you know, many different, like, registries of MCP servers that even guys from Master, our friends, created, like, registry of MCP server registries, right?

00:07:06.120 | Just to make the sense of it, right?

00:07:08.120 | And, obviously, Anthropic is also working on their own registry, and I think Google's A2A, they have, like, a DNS-based protocol with, like, well-known .agents.json way to, you know, publish the services through DNS.

00:07:23.120 | So, basically, there is, like, you know, so many different servers you can now use from the agents, right?

00:07:29.120 | So, does it mean, like, so many tools now support MCP, so does it mean, like, the agents can discover and access any of them on their own, right?

00:07:40.120 | Well, not really, because to use those services, your agents still need to have, like, API tokens to those services, right?

00:07:49.120 | So, even, let's say, if you use Zapier MCP, that provides access to, like, 5,000 apps they have in their marketplace,

00:07:56.120 | you still need to connect those individual apps to your services, right?

00:08:00.120 | You know, like, GitHub or Slack or, you know, whatever.

00:08:03.120 | So, Zapier, on its own, is not able to provide access to the third-party services.

00:08:08.120 | You still need to, as a user, to facilitate it.

00:08:11.120 | So, that actually means that the agents are not able to, like, find counterparts or, like, other agents or other tools to interact with on their own.

00:08:23.120 | They are still depending on the human developer who actually build the system, right?

00:08:27.120 | Who kind of, like, give those agents access to different tools, right?

00:08:32.120 | And if those agents are, you know, to replace all the people and all the jobs, right, they need to be able to find services to interact with.

00:08:41.120 | They can't just, like, you know, do that.

00:08:43.120 | Like, it's, like, a basic thing that, like, anyone of us can do, right?

00:08:48.120 | Like, to find service and purchase it, right?

00:08:50.120 | So, I argue that, like, unless the agents are able to do that, we will not be able to reach, you know, some higher level of intelligence of these agentic systems and behaviors, basically, if the agents cannot purchase services, right?

00:09:06.120 | So, how can we solve this problem, right?

00:09:10.120 | So, first, like, sort of, like, a naive approach would be let the agents subscribe themselves to target services, right?

00:09:16.120 | So, basically, in a way, like, agents could have, like, email, maybe a credit card.

00:09:20.120 | They could, like, fill, you know, the subscription flow, maybe solve the captcha, you know, create an account and so on.

00:09:26.120 | But you see, it's not very practical, right?

00:09:28.120 | I mean, it's, you know, well, they might need to also have to phone number and so on.

00:09:33.120 | And quite often, the services actually need to have, like, a real person behind the account, right?

00:09:38.120 | So, basically, this wouldn't really work, right?

00:09:41.120 | So, second solution is central identity and payments provider.

00:09:46.120 | There are, like, a couple of companies pursuing now that, like, there would be, like, a central authority where you can charge money.

00:09:51.120 | And then the agents can use that, you know, to buy services and provide them with their identity, right?

00:09:56.120 | For example, Vertifier, Coinbase is now pushing their X402 standard.

00:10:01.120 | Stripe is working on this and MasterCard and Visa, too, right?

00:10:05.120 | So, I think this is going to happen eventually.

00:10:07.120 | But running or launching a new payment system is extremely complicated, right?

00:10:12.120 | Because you are facing, like, this chicken and egg problem of marketplaces, right?

00:10:15.120 | I think PayPal had to pay, like, $100 million per month just to buy the market.

00:10:20.120 | And, like, launching credit cards in the '70s was, like, incredible challenge, basically.

00:10:25.120 | Because nobody was accepting those cards, so why would people use them and so on, right?

00:10:28.120 | So, I think this will happen, but it will be a long process, basically, to establish this, right?

00:10:33.120 | So, let me offer the third approach.

00:10:35.120 | And it's, like, through a centralized marketplace of MCP services, like, if I store, basically,

00:10:41.120 | where you just need one API token or one authentication, one account, to get access to all the other services.

00:10:47.120 | And, basically, it works the way that the developers who publish these tools, these actors,

00:10:52.120 | actually, they provide their credit card and their account to the third-party service,

00:10:56.120 | and, basically, publish it, add monetization to it, like, how much does it cost to call the service?

00:11:01.120 | And then they are, basically, the owner of the service.

00:11:04.120 | And now they publish it on our marketplace, and suddenly it becomes available to the whole ecosystem of tools.

00:11:09.120 | And this way, actually, we can scale it rapidly and, actually, even without the target services knowing, right?

00:11:15.120 | So, basically, this way, the actor can run the code itself or wrap an external API or just publish an external MCP server,

00:11:22.120 | because the MCP servers, they can be actually nested.

00:11:25.120 | You can have, like, one parent server that provides actions or tools of the, like, nested MCP servers, right?

00:11:31.120 | So, that's another, like, cool feature of FMCP.

00:11:34.120 | You can really build this sort of ecosystem, you know, if you can facilitate the payments and monetization, right?

00:11:39.120 | So, actor charges the user, and then its developer gets the money and pays for the external service,

00:11:46.120 | and anyone can publish such an actor even without the target service knowing, right?

00:11:51.120 | So, time for demo.

00:11:52.120 | It's not live demo because the internet is super flaky here.

00:11:56.120 | So, what you can see here is a cloud for desktop that has access to API MCP server.

00:12:02.120 | There is, like, 18 tools available now.

00:12:04.120 | And I'm asking, like, what is the venue of AI Engineer World's Fair in San Francisco?

00:12:09.120 | It's possible to use API Actors.

00:12:10.120 | So, you can see it searches the actors for a tool that can answer this question.

00:12:17.120 | It will find a tool or actor called RagWeb Browser.

00:12:20.120 | And so, it's like a Google search with, you know, fetch data.

00:12:25.120 | So, basically, it asks the query, like, what is the venue, and so on.

00:12:29.120 | And then it parses the resulting page.

00:12:31.120 | So, we can see it found, like, SF Marriott Marquis.

00:12:35.120 | That seems all correct, right?

00:12:37.120 | So, now, let's use an actor for scraping Twitter.

00:12:41.120 | So, this actor is not available in the context.

00:12:46.120 | So, the agent doesn't know how to use it.

00:12:48.120 | So, it searches actors on our store and finds an actor that can scrape Twitter, right?

00:12:54.120 | So, it calls add actor, which is, like, a tool that adds a new tool to the context.

00:13:00.120 | Actually, Claude is very verbose, describing a lot of things about it.

00:13:03.120 | And, actually, there is, like, small bug still in Claude Desktop that you need to, like, disable and enable a tool so that the tool list refreshes and then the tools become available.

00:13:12.120 | I'm sure it's going to be fixed in the next release.

00:13:15.120 | And, now, let's use that actor to get the last tweet of AI Engineer Conference.

00:13:20.120 | All right?

00:13:21.120 | So, it calls the actor on Appify.

00:13:23.120 | It knows the Twitter handle, probably from the website.

00:13:28.120 | And, now, you can see that it found the result, and the last tweet from it this morning was something about workshops.

00:13:36.120 | That seems about right.

00:13:38.120 | So, now what?

00:13:39.120 | So, we have seen how we can use existing tools in our store.

00:13:45.120 | But, like, let's say, one of our competitors, a company called BrowserBase.

00:13:51.120 | Hey, Paul, if you're here.

00:13:53.120 | They certainly haven't published, you know, an actor in our store.

00:13:58.120 | But, we did.

00:13:59.120 | So, we created an account on BrowserBase, added our API token there, and published, like, basically, their MCP server on our store without actually them even knowing.

00:14:09.120 | And, now, anybody can actually use BrowserBase MCP through Appify's ecosystem, right?

00:14:15.120 | Even without them having to do anything or knowing about it, right?

00:14:18.120 | So, now, let's use BrowserBase to fill in the email subscription form on the AI engineer website.

00:14:25.120 | Fill email, janetappify.com.

00:14:28.120 | And, now, let's see what happens, right?

00:14:30.120 | And, actually, we'll see that the agent will actually call BrowserBase MCP through, you know, an actor published, you know, by us on our Appify store, and perform the actions on the web, right?

00:14:43.120 | And, actually, this way, we can easily, like, bring a lot of existing MCP servers to our store, and, you know, expand the ecosystem rapidly without, you know, having to ask for, you know, cooperation of the third parties, right?

00:14:57.120 | So, that's actually what we're doing now.

00:14:59.120 | We want to scale this marketplace rapidly.

00:15:02.120 | And, now, okay, so, now it's evaluating, you know, the screenshots, looking for the field, and so on, you know.

00:15:09.120 | And, eventually, it will manage to fill the form, and basically succeed in the task, right?

00:15:16.120 | I can maybe skip this to save time.

00:15:20.120 | It takes some time to, basically, for the agent to find the form, and so on.

00:15:27.120 | But, yeah.

00:15:29.120 | It succeeded.

00:15:30.120 | It completed the email subscription, and this way, you basically see that you can plug our ecosystem of actors into any AI agents that actually support tool discovery, right?

00:15:46.120 | So, this means, now, anyone can publish tools or, you know, agents on Appify store, and monetize them, and immediately get access, you know, to all the AI clients that already, like, integrated Appify and all the ecosystem of tools, right?

00:16:08.120 | And, actually, people can make money on it.

00:16:11.120 | Like, just last month, we paid more than a quarter million dollars to our creators.

00:16:15.120 | And, actually, this number is growing rapidly, you know.

00:16:18.120 | Overall, the actors generate more than one and a half million dollars per month now.

00:16:22.120 | We have, like, one million monthly visitors to the whole ecosystem.

00:16:25.120 | And, now, we're in the process of, like, scaling this ecosystem.

00:16:28.120 | So, if you're looking for ways to monetize your tools or agents, you know, just talk to us and publish or publish or publish a store and get access to this ecosystem of developers and this visibility.

00:16:43.120 | And, there are some open questions, obviously, that remain.

00:16:48.120 | So, will this autonomous tool discovery provide real value?

00:16:53.120 | I mean, like, everybody who builds agentic systems knows that, you know, like, making sure that the system works as expected is tricky, right?

00:17:02.120 | Even if it's fixed.

00:17:03.120 | So, if we add this, like, you know, variables that, like, well, if the agents can discover new tools, will it actually work?

00:17:12.120 | Well, currently, it might be a bit flaky, right?

00:17:14.120 | I think we're still fairly early.

00:17:17.120 | But, as the models get better, I think, even with the discovery, suddenly, the agents will be able to provide, you know, valuable and reliable results, basically, right?

00:17:31.120 | So, this remains to be seen, but I'm optimistic that, like, as the elements will get better, we'll actually get there, that the tool discovery will actually provide real value.

00:17:40.120 | Well, there's a big question of, like, how can agents trust tools or other tools?

00:17:45.120 | Oh, sorry, or each other, right?

00:17:46.120 | We know it, like, you only interact with people you trust.

00:17:49.120 | So, how can agents do that?

00:17:50.120 | We'll see.

00:17:51.120 | And, can autonomous agent interaction enable AGI?

00:17:54.120 | Well, we'll see.

00:17:55.120 | Thank you very much for your attention.

00:17:57.120 | And, feel free to try it.

00:17:59.120 | mcpu.wi-fi.com.

00:18:00.120 | mcpu.wi-fi.com.

00:18:03.120 | you

00:18:03.620 | you

00:18:05.680 | you

The rise of the agentic economy on the shoulders of MCP — Jan Curn, Apify

Chapters