Jan Čurn is the co-founder and chief executive of Apify, a Prague-rooted platform for web scraping, browser automation, and software that AI agents can discover and use. He is building an agentic economy where specialized tools can be found, authorized, purchased, and combined without requiring every integration to be assembled by hand.
From web crawlers to autonomous software
In 2015, Čurn and Balada launched Apify through the Y Combinator Fellowship in Mountain View. Their product used headless browsers and JavaScript to help developers extract information from increasingly dynamic websites. They returned to the Czech Republic in 2016, raised seed funding, and expanded into web scraping and browser automation.
Čurn introduced an open-source JavaScript scraping library in a 2018 article announcing Apify’s release of the Apify SDK. His 2019 marketplace blueprint imagined developers publishing, selling, and running reusable cloud software—a vision that now underpins Apify’s approach to AI agents.
Its core building blocks are Apify Actors: Docker-based programs with defined inputs and outputs that can collect web data, automate browsers, or process information. Independent developers publish and monetize Actors alongside Apify’s own tools, while APIs, command-line interfaces, and Python and TypeScript integrations make them accessible to other software.
Dynamic tool discovery: Loading an entire marketplace into a model’s context is impractical. Čurn uses the Model Context Protocol to let agents locate relevant tools and add their interfaces only when needed, reducing context overhead and making large software catalogs usable.
Agent identity and payments: Discovering a service does not provide credentials, authorization, or billing. Čurn’s marketplace model allows tool developers to manage external integrations and payment while agents connect through a shared account and interface.
Nested MCP integrations: An Actor can expose another service’s capabilities through Apify’s marketplace. Čurn demonstrated the pattern with Browserbase, connecting browser automation to an agent through an intermediary integration.
Operationally useful interoperability: Čurn’s article introducing mcpc, a universal MCP command-line client, describes addressing repeated authentication, fragmented configuration, and inefficient tool loading. He also announced MCP connectors, which provide authenticated Actor access to external applications through manually assigned connectors, tool restrictions, and an Apify-managed proxy that keeps credentials outside the Actor runtime. These announcements establish his publication role without implying sole software authorship.
Čurn’s personal Apify account lists 10 public Actors. His public positions combine enthusiasm for MCP with skepticism about premature machine-payment infrastructure: at AI Engineer World’s Fair in 2026, he defended MCP while questioning x402’s readiness. He also considers collective intelligence emerging from interacting agents plausible, while treating trust, reliability, and autonomous commerce as unresolved engineering problems.
Dynamic tool discovery lets agents find new capabilities, but acquiring services also requires credentials and payment. Apify’s marketplace shows how those pieces can fit together.
How does intelligence emerge from connected parts?
Jan Curn explains what Apify’s agent-payment integration exposed: verified signatures do not reserve money, fixed prices do not fit metered jobs, and payment protocols can collide with authentication. Prepayment makes the system usable; escrow and batch settlement offer a promising direction.
Verification does not secure funds in the x402 flow described here. Running costly work after settlement protects sellers from spending that occurs between authorization and collection.
Mandatory initial status codes complicate combining x402 payments with MCP authentication. Separate hostnames work around the conflict; Curn proposes allowing payment requirements through headers without requiring 402.
Up To permits a variable charge within a ceiling but leaves the spending window open. Apify’s charge-and-refund workaround adds a second blockchain transaction and requires buyers to trust the refund.
Batch Settlement combines upfront escrow with off-chain micropayments and aggregated on-chain settlement. Apify had not yet implemented it at talk time.
Apify’s Agent General Interface separates purchasing prepaid access from executing jobs through the established API or MCP. Its documented prepaid credit terms differ from the talk’s earlier refund workaround.
▶ Watch ↗