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

So let me start with a question: how does intelligence emerge in biological systems? Right? Well, it's through neurons,right? When neurons are born, they are just like individual cells, but over time they grow their axons and dendrites and establish connections with other cells or other neurons, and actually learn how to communicate in order to pursue their own interests, basically, like to get nutrients and so on.

And over time they learn how to communicate with each other and with other cells to get nutrients and basically thrive,right? And this collective behavior, if you zoom out and look at a really large number of them, it's something we call intelligence,right?

So it's like emergent behavior of smaller individual units that pursue their own interests. So how does intelligence emerge in the markets,right? People always talk about markets like, well, market thinks that, market reacted to this, and so on. And in some way, markets are more intelligent than individual participants of the market,right?

And it's their mutual interaction of these individual members of the market, basically, who pursue their own interests and communicate and establish new interactions with others, where some sort of collective intelligence, which is bigger than the sum of different parts, emerges,right?

So how does intelligence emerge in companies? Well, this one is provocative: through Slack,right, where people interact and pursue their own interests in the company, and altogether the company, well, sometimes becomes more intelligent than the individual employees of the company.

And so this leads to my final question: so how does, or how will, the general intelligence emerge in computing systems,right? And there is a lot of talk about AGI and

ever larger models exhibiting superintelligent behavior, 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 altogether exhibiting something which we can call general intelligence.

And thanks to MCP, we finally have this missing part that allows the agents to communicate with each other and really create a fabric or agentic mesh where they can talk together. So hello everyone, my name is Jan Curn, I'm the founder of Apify, and I'm going to talk about the rise of the agentic economy on the shoulders of MCP, basically, economy where agents can find counterparts to interact with and purchase services from businesses or tools or other agents,right?

So like B2A and B2B, sorry, and A2A. Allright, so before I start, let me just introduce quickly Apify. Apify is a marketplace of 5,000 tools called Actors, and historically we come from the web scraping industry,right? So most of these Actors are data extraction tools that allow you to get data from social media, from search engines, data for AI for building RAG pipelines, data from web for lead generation, and so on.

But also there are other tools, like data processing tools and so on. So altogether there's about 5,000 of them, and some of them are built by Apify, some are built by our community of creators who actually make money on it,right?

So it's like a marketplace of software creators, if you will,right? So Actors are self-contained pieces of software based on Docker with well-defined input and output,right? And basically they represent a new way how to ship software and publish it and integrate to other systems,right?

So for example, Google Maps Scraper, it's a quite popular Actor from our store. It can extract data from Google Maps,right? More data than the Google Places API provides,right? Well, there is creator of the Actor description, different stats, and so on, something you would expect from a normal marketplace.

And actually, thanks to the way Actors are built, it's actually super easy to integrate Actors from other systems,right? So for example, we have SDKs for TypeScript, for Python, for OpenAPI, for CLI, which you can call them from Terminal.

And it's only because they are well-defined units of software with input and output,right? Also, we have integrations with workflow automation tools like Make, Zapier, Clay, and many others, so to make it really easy to call Actors from these systems,right?

But obviously now we also have MCP integration, which makes it possible to call Actors from AI agents or AI workflows. And the way it works, actually, is

the agent just needs an API key or OAuth workflow on an account on Apify, and then through our MCP server, basically, it can interact or call any of those 5,000 Actors on our marketplace,right? And actually, this only became possible thanks to, I would say, the killer feature of MCP, which is the tool discovery,right?

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, Claude for Desktop added support for tool discovery. 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?

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

So we really want to provide the tools only as needed. And that's only possible through tool discovery, which I think is really the main thing that will actually make MCP really

the huge differentiator from OpenAPI, for example,right? So MCP actually quickly became a standard for agentic interaction. This is Google Trends data showing that MCP is basically dominating the space compared to OpenAPI or A2A from Google,right? And actually, I think MCP already became the standard for agentic interaction.

And it became so popular that currently there are many different registries of MCP servers that even guys from Master, our friends, created a registry of MCP server registries,right? Just to make the sense of it,right? And obviously, Anthropic is also working on their own registry.

And I think Google's A2A, they have a DNS-based protocol with a well-known .agents.json way to publish the services through DNS. So basically, there are so many different servers you can now use from the agents,right? So does it mean that so many tools now support MCP, so does it mean the agents can discover and access any of them on their own,right?

Well, not really, because to use those services, your agents still need to have API tokens to those services,right? So even, let's say, if you use Zapier MCP, that provides access to 5,000 apps they have in their marketplace, you still need to connect those individual apps to your services,right?

Like GitHub or Slack or whatever. So Zapier, on its own, is not able to provide access to the third-party services. You still need to, as a user, to facilitate that. So that actually means that the agents are not able to find counterparts or other agents or other tools to interact with on their own.

They are still depending on the human developer who actually built the system,right? Who kind of gave those agents access to different tools,right? And if those agents are to replace all the people and all the jobs,right, they need to be able to find services to interact with.

They can't just do that. It's like a basic thing that anyone of us can do,right? Like to find services and purchase it,right? So I argue that unless the agents are able to do that, we will not be able to reach some higher level of intelligence of these agentic systems and behaviors, basically, if the agents cannot purchase services,right?

So how can we solve this problem,right? So first, sort of a naive approach would be to let the agents subscribe themselves to the target services,right? So basically, in a way, agents could have email, maybe a credit card, they could fill the subscription flow, maybe solve the CAPTCHA, create an account, and so on.

But you see, it's not very practical,right? I mean,

well, they might also have to phone number and so on, and quite often the services actually need to have a real person behind the account,right? So basically, this wouldn't really work,right? So the second solution is a central identity and payments provider.

There are a couple of companies pursuing now that there would be a central authority where we can charge money, and then the agents can use that to buy services and provide them with their identity,right? For example, Verifier, Coinbase is now pushing their X402 standard.

I think Stripe is working on this, and Mastercard and Visa too,right? So I think this is going to happen eventually, but launching a new payment system is extremely complicated,right? Because you're facing this chicken and egg problem of marketplaces,right?

I think PayPal had to pay like $100 million per month just to buy the market, and launching credit cards in the '70s was an incredible challenge, basically, because nobody was accepting those cards, so why would people use them, and so on,right?

So I think this will happen, but it will be a long process, basically, to establish this,right? So let me offer the third approach, and it's like through a centralized marketplace of MCP services, like Apify store, basically, where you just need one API token or one authentication, one account to get access to all the other services.

And basically, it works the way that the developers who publish these tools, these Actors, actually, they provide their credit card and their account to the third-party service and basically publish it, add monetization to it, like how much does it cost to call this service, and then they are basically the owner of the service, and now they publish it on our marketplace, and suddenly it becomes available to the whole ecosystem of tools.

And this way, actually, we can scale it rapidly and actually even without the target services knowing,right? So basically, this way, the Actor can run the code itself or wrap an external API or just publish an external MCP server, because the MCP servers, they can be actually nested.

You can have one parent server that provides actions or tools of the nested MCP servers,right? So that's another cool feature of MCP. You can really build this sort of ecosystem if you can facilitate the payments and monetization,right? So Actors charge the user, and then its developer gets the money and pays for the external service, and anyone can publish such an Actor even without the target service knowing,right?

So time for a demo. It's not a live demo because the internet is super flaky here. So what you can see here is Claude for Desktop

that has access to Apify MCP server. There are like 18 tools available now, and I'm asking, what is the venue of AI Engineer Awards Fair in San Francisco? If possible, use Apify Actors. So you can see it searches the Actors for a tool that can answer this question.

It will find a tool or Actor called RAG Web Browser, and so it's like a Google search with fetch data. So basically, it asks the query, like what is the venue, and so on, and then it parses the resulting page.

So we can see it found SF Marriott Marquise. That seems all correct,right? So now let's use an Actor for scraping Twitter. So this Actor is not available in the context, so the agent doesn't know how to use it.

So it searches Actors on our store and finds an Actor that can scrape Twitter,right? So it calls addActor, which is like a tool that adds a new tool to the context. Actually, Claude is very verbose, describing a lot of things about it.

And actually, there are small bugs still in Claude Desktop that you need to disable and enable a tool so that the tool list refreshes, and then the tools become available. I'm sure it's going to be fixed in the next release.

And now let's use that Actor to get the last tweet of the AI Engineer conference, allright? So it calls the Actor on Apify. It knows the Twitter handle, probably from the website. And now you can see that it found the result, and the last tweet from this morning was something about workshops.

That seems aboutright. So now what? So we have seen how we can use existing tools in our store, but let's say

one of our competitors, a company called Browserbase. Hey, Paul, if you're here. They certainly haven't published an Actor in our store, but we did. So we created an account on Browserbase, added our API token there, and published basically their MCP server on our store without actually them even knowing.

And now anybody can actually use Browserbase MCP through Apify's ecosystem,right? Even without them having to do anything or knowing about it,right? So now let's use Browserbase to fill in the email subscription form on the AI Engineer website. Fill email yannetapify.com, and now let's see what happens,right?

And actually, we'll see that

the agent will actually call Browserbase MCP through an Actor published by us on Apify's store and perform the actions on the web,right? And actually, this way, we can easily bring a lot of existing MCP servers to our store and expand the ecosystem rapidly without having to ask for cooperation of the third parties,right?

So that's actually what we're doing now. We want to scale this marketplace rapidly. And now, okay, so now it's evaluating the screenshots, looking for the field, and so on, and eventually, it will manage to fill the form and basically succeed in the task,right?

I can maybe skip this to save time. It takes some time to basically for the agent to find the form and so on, but yeah, it succeeded. It completed the email subscription, and

this way, you basically see that you can plug our ecosystem of Actors into any AI agents that they actually support tool discovery,right? Allright, and

so this means now anyone can publish tools or agents on Apify's store and monetize them and immediately get access to all the AI clients that already integrate with Apify and all the ecosystem of tools,right? And actually, people can make money on it.

Just last month, we paid more than a quarter million dollars to our creators, and actually, this number is growing rapidly. Overall, the Actors generate more than one and a half million dollars per month now. We have one million monthly visitors to the whole ecosystem, and now we're really in the process of scaling this ecosystem.

So if you're looking for ways to monetize your tools or agents, just talk to us and publish or publish Actor on Apify's store and get access to this ecosystem of developers and this visibility. And there are some open questions, obviously, that remain.

So will this autonomous tool discovery provide real value? I mean, everybody who builds agentic systems knows that making sure that the system works as expected is tricky,right? Even if it's fixed. So if we add these variables that, well, if the agents can discover new tools, will it actually work?

Well, currently, it might be a bit flaky,right? I think we're still fairly early, but as the models get better, I think

even with the discovery, suddenly the agents will be able to provide valuable and reliable results, basically,right? So this remains to be seen, but I'm optimistic that as the LMs will get better, we'll actually get there that the tool discovery will actually provide real value.

Well, there's a big question of how can agents trust tools or other tools, sorry, or each other,right? We know that you only interact with people you trust, so how can agents do that? We'll see. And can autonomous agent interaction enable AGI?

Well, we'll see. Thank you very much for your attention, and feel free to try it at mcp.apify.com.