The MCP Kingmaker? How Czech Apify Became the Stripe for AI Agents
Apify (Prague, est. 2016) has repositioned from a traditional web-scraping platform into an AI-native marketplace — and the timing couldn't be better.
The core thesis: Apify spent a decade building a marketplace of ~5,000 "actors" (Docker-based cloud apps) for web scraping & automation. That actor abstraction — standardized inputs, outputs, SDKs, and integrations — turns out to be almost perfectly shaped for the emerging AI agent ecosystem. They're now wrapping external MCP (Model Context Protocol) servers as Apify actors, positioning themselves as the payment and discovery layer for autonomous AI agents. The pitch: agents need a single API key to find, use, and pay for tools, and Apify is building that middleware.
The numbers back it up:
- Web traffic has gone parabolic: ~650k visits in early 2024 → ~2.6M visits recently (~4x in under 2 years)
- Revenue trajectory: $5.3M (2021) → $6.6M (2022) → $7.7M (2023), with the marketplace now generating $1M+/month and paying creators ~$250k/month
- Headcount surging to ~196 people (sharp recent spike)
- Dealroom signal score: 90 (very high)
- Just $4.5M total funding (last round: €2.8M Early VC from J&T Ventures, Apr 2024) — capital efficient
Key catalyst: The MCP wave. As Anthropic's Model Context Protocol gains adoption, the bottleneck shifts from "can agents use tools?" to "how do agents discover, authenticate, and pay for tools?" Apify is betting they're the answer — an open marketplace layer where agents transact via a single API key while actor creators handle upstream billing. They're demoing wrapped competitor tools (including Playwright MCP) running through their platform.
Watch for: A likely Series A in the near term given the traffic/revenue inflection and relatively small funding base. Open questions remain around reliability of autonomous tool discovery, trust/scam prevention between agents, and whether the marketplace model creates sufficient lock-in.
