Something shifted on Apify in early 2026. They added a new category to the Store: MCP_TOOLS. If you're building actors and you haven't noticed this yet, you're about to miss the biggest distribution shift since pay-per-event pricing launched.
I'm Ryan, the developer behind ApifyForge. We've shipped over 80 MCP intelligence servers on the Apify Store under the username ryanclinton. That's not a typo. Eighty-plus servers, live right now, covering financial crime screening, counterparty due diligence, cybersecurity threat assessment, supply chain risk, regulatory intelligence, and about a dozen other domains. We started building these before Apify even had an official MCP category. And now that the category exists, the timing couldn't be better.
Here's what's actually going on, why it matters, and why most Apify developers are going to be late to this.
What is an MCP server on Apify?
An MCP server is a tool server that speaks the Model Context Protocol, an open standard created by Anthropic in late 2024 that lets AI assistants like Claude, ChatGPT, and Cursor call external tools directly. Instead of copying data into a chat window, the AI connects to an MCP server and runs tools on its own — querying databases, pulling records, scoring risk, returning structured results.
On Apify, an MCP server runs in standby mode. It stays warm, responds to tool calls in seconds, and charges per event. No idle compute costs. No monthly subscriptions. You pay when the AI actually uses the tool. Apify's infrastructure handles the scaling, the hosting, the uptime — the developer just builds the intelligence layer.
According to Anthropic's MCP specification docs, the protocol has seen adoption across Claude Desktop, Cursor, Windsurf, Continue, and dozens of other AI clients since its November 2024 launch. The MCP client directory now lists dozens of AI tools with native support, from Claude Desktop to VS Code to Cursor. The growth curve hasn't slowed down.
Why are MCP servers a bigger deal than regular actors?
Regular Apify actors are great. I've built over 170 of them. But they have a friction problem: somebody has to go to the Apify console, configure the input, hit run, wait for results, then do something with the output. That workflow made sense when humans were the primary interface.
MCP servers remove that friction entirely. An AI assistant discovers the server, reads its tool descriptions, and calls the right tool with the right parameters. No console. No manual configuration. The AI figures it out from the schema.
That matters because the buyer is changing. In 2024, the typical Apify customer was a developer or a growth marketer running scrapers. In 2026, the typical customer is an AI agent that needs real-time data. Those agents need tools. MCP is how they find and use them.
Here's the blunt version: if your actor can't be called by an AI agent, you're building for a shrinking audience.
How many MCP servers are on the Apify Store right now?
As of March 2026, the Apify Store's MCP_TOOLS category is still new. Most developers haven't built any MCP servers yet. ApifyForge has over 80 live — which, tbh, is a wild number. We're the most prolific MCP publisher on the platform by a significant margin.
That head start wasn't an accident. We started building MCP intelligence servers in late 2025, months before Apify launched the official category. The bet was simple: AI assistants would need domain-specific tools, not generic scrapers. A compliance officer using Claude doesn't want a "web scraper." They want a tool that screens an entity against sanctions lists, PEP databases, and adverse media — and returns a risk score they can act on.
So that's what we built. Eighty-plus of them.
What domains do ApifyForge MCP servers cover?
ApifyForge MCP servers span nine major intelligence domains: financial crime and AML screening, counterparty and corporate due diligence, cybersecurity and attack surface analysis, ESG and supply chain risk, regulatory and compliance monitoring, geopolitical risk assessment, M&A target intelligence, critical infrastructure analysis, and competitive intelligence.
Each server isn't a thin wrapper around a single API. That would be pointless — you can call a single API yourself. The value is in orchestration. A single tool call to the Counterparty Due Diligence MCP hits multiple public data sources in parallel, cross-references the results, and returns a scored assessment. The AI assistant gets a structured answer, not a pile of raw JSON it has to figure out.
The Financial Crime Screening MCP does the same thing for AML workflows. The ESG Supply Chain Risk MCP does it for sustainability and supply chain due diligence. The Entity Attack Surface MCP does it for cybersecurity. Different domains, same pattern: multiple sources, parallel execution, domain-specific scoring.
We didn't build 80+ toy demos. We built production intelligence tools that happen to speak MCP.
What does pay-per-event pricing mean for MCP servers?
Pay-per-event (PPE) pricing means the user pays a fixed price per tool call — not per minute of compute time, not per monthly subscription. ApifyForge MCP servers typically charge between $0.05 and $0.50 per tool call, depending on how many data sources the server orchestrates and how much processing each call requires.
This pricing model is a natural fit for AI agents. An agent doesn't know in advance how long a task will take or how much compute it needs. With PPE pricing, the cost is predictable: call the tool, get the result, pay the fixed rate. Apify's paid actors documentation explains how this works at the platform level — the developer sets the event price, Apify handles billing, and the user sees a clear per-call cost.
Compare that to building in-house. A single compliance screening tool costs six figures to build internally when you factor in development, API licensing, maintenance, and compliance validation. An ApifyForge MCP server that does the same screening costs $0.15 per call. Run it a thousand times and you've spent $150.
The economics are lopsided. That's the point.
Why hasn't everyone built MCP servers on Apify yet?
Honestly? Most Apify developers are still thinking in the old model. Build a scraper, put it in the Store, hope someone finds it. MCP servers require a different approach. You need to understand how AI assistants discover and invoke tools. You need to design tool schemas that an LLM can reason about. You need to handle standby mode, which behaves differently from standard actor runs. And you need domain expertise to build something that returns intelligence, not just data.
That last part is the real barrier. Anyone can scrape a website. Not everyone can build a financial crime screening workflow that knows which sanctions lists matter, how to handle fuzzy name matching across multiple alphabets, and what a risk score should actually look like for a given entity type.
The MCP servers that will win aren't the ones built fastest. They're the ones built by people who actually understand the domain. We've spent months studying compliance screening workflows, corporate due diligence processes, and cybersecurity assessment frameworks. That domain knowledge is baked into every scoring algorithm, every data source selection, every output schema.
How do AI assistants find MCP servers on Apify?
AI clients like Claude Desktop, Cursor, and Windsurf use MCP server URLs to connect. The user adds the server's standby URL to their client configuration, and the AI assistant automatically discovers the available tools through the MCP protocol's built-in tool listing mechanism. No plugins to install, no OAuth flows, no marketplace approvals.
Apify's Store acts as a discovery layer. When someone searches for "sanctions screening" or "due diligence" on the Apify Store, they find the MCP server, copy the connection URL, and add it to their AI client. The entire setup takes about 30 seconds, which I wrote about in our getting started guide.
This is a very different distribution model than traditional SaaS. There's no sales cycle. No demo call. No onboarding flow. An AI assistant just... connects and starts calling tools. The M&A Target Intelligence MCP gets used by investment analysts running due diligence through Claude. They didn't fill out a form or talk to sales. They added a URL.
The window is open right now
I'll be direct about this. The MCP category on Apify is new. Most developers haven't started building servers yet. The ones who move now will own the search real estate, build the usage history, and accumulate the reviews that make it hard for latecomers to compete.
We saw this exact pattern play out with regular actors. The early publishers — the ones who shipped quality actors in 2022 and 2023 — still dominate the Store rankings in 2026. Search position compounds. Usage history compounds. Reviews compound.
MCP servers are following the same curve, but earlier. They're the connective tissue between AI agents and the real world.
ApifyForge is already there. Eighty-plus servers. Nine intelligence domains. Production-grade scoring algorithms. PPE pricing that makes sense for agent workflows. You can browse the full catalog at apifyforge.com/actors/mcp-servers or find us in the Apify Store under ryanclinton.
MCP is where Apify is going. We're already there.
Last updated: March 2026
Top comments (0)