“Developers see this as the future”: Pilot Protocol launches to power the agent economy
The New Stack Adrian Bridgwater
Pilot Protocol launched a marketplace where AI agents hire other AI agents and tools, paying from their own wallets. It's basically an app store built for bots, not humans — and 250,000 agents are already using it.
Based on reporting by The New Stack, Adrian Bridgwater — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Software agents have always been built like lone employees: one owner, one machine, no way to talk to peers unless a human bothers to wire up an API or point them at an MCP server. Pilot Protocol wants to break that isolation. The company came out of stealth this week with a platform that gives agents their own network address, a bit like a parallel internet just for bots, where they can find each other plus tools and apps built to extend what they can do.
CEO Razvan Roman says the demand didn't come from a sales pitch, it came from the agents themselves. Install Pilot with one line of code, he says, and an agent can go hunting for currency data, traffic info, legal answers, GitHub packages, whatever it needs to finish its assigned job. Roman claims every developer he talks to wants in, mostly because shipping a product currently means negotiating with humans, and nobody building autonomous systems wants to keep doing that. Pilot wraps a developer's app and slots it into a curated store, and Roman says the network already counts 250,000 agents generating two billion requests a day, most of it happening without their human owners even noticing. Growth has hit 10% daily at points, with 16,000 agents joining in a single 24-hour stretch.
Every agent on Pilot gets a wallet. Instead of app makers buying ads to reach humans, distribution happens inside the network itself, agents shop for tools based on merit and pay per use, sometimes earning money back if they read an ad. Roman is emphatic that Pilot doesn't set prices; it just runs the store and takes a commission when an agent pays for something, App Store style. That's a notable design choice given how much of today's SaaS runs on annual contracts, a model that makes little sense for an agent that only needs a tool for ten minutes. Bain estimates agent-driven commerce could hit $300-500 billion in the US by 2030, and Pilot's own team is betting on a trillion agents online within five years.
One early example: smolmachines, a company managing Linux virtual machines, listed on Pilot's store and picked up 3,000 agent installs in days with zero marketing spend. Nobody drove those installs, according to CTO BinBin H, the agents just showed up and started provisioning machines on their own. Under the hood, Pilot runs on UDP rather than the connection-heavy TCP, streaming data without waiting on delivery confirmations, and the install process is built with zero dependencies so agents don't accidentally drag home malware while shopping.
As for the name, Pilot rather than Autopilot: Roman's logic is that autopilots follow fixed, deterministic paths, while pilots make judgment calls and change course when something better shows up. Fitting, given the whole pitch is agents making their own decisions about who to talk to and what to pay for.
My take — AI-written commentary, not fact-checked reporting
An economy where the customers are bots buying tools from other bots with wallets nobody's watching is exactly the kind of thing that sounds efficient right up until it isn't, and the fact that most of this happens without owners' knowledge should worry people more than the growth numbers excite them. I'm all for killing SaaS's annual-contract nonsense, usage-based pricing for agents is obviously correct, but 'agents just show up and onboard themselves' is a sentence that should come with a security audit attached, not a press release. This is the agent hype cycle doing what it always does: shipping the infrastructure before anyone's worked out who's liable when an autonomous wallet gets drained.
Read more about this at: The New Stack