AI agents can write code, book meetings, browse the web, and complete tasks with little human input. One thing they still struggle to do is find and work with other agents on their own. San Francisco startup Pilot Protocol believes that missing layer is about to become one of AI’s biggest infrastructure opportunities, and investors just backed that bet with fresh capital.
Pilot Protocol emerged from stealth on Monday with $4.5 million in seed funding led by Version One Ventures. The round included participation from Precursor Ventures, Night Capital, Todd & Rahul Capital, plus angel investors Lenny Rachitsky and Ben Tossell. The company says it is building a network where AI agents can discover one another, establish trust, exchange services, and eventually make payments without relying on human setup.
The announcement arrives at a time when the AI industry is shifting from standalone chatbots to autonomous software agents that can complete increasingly complex workflows. Many AI startups are racing to build agents. Far fewer are focused on the infrastructure that lets those agents communicate across different systems.
Pilot Protocol launches AI agent network with $4.5M seed led by Version One Ventures
Pilot’s pitch is straightforward. Instead of treating every AI agent as an isolated application, the company assigns each agent its own network identity and gives it a way to discover and interact with other agents directly. In practice, that means an agent searching for a capability such as scheduling, coding, research, or cloud computing could locate another specialized agent and hand off work without human intervention.
The company says developers can connect an agent to the network with a single installation command. From there, agents can discover services, verify trusted applications, and communicate across the network.
Pilot claims adoption has happened with little traditional marketing. The startup says roughly 250,000 agents are already connected to its network, generating about two billion requests every day. It says growth reached as much as 10% per day during its early expansion, with roughly 16,000 new agents joining within a 24-hour period at one stage.
The company argues that many connected agents eventually begin using Pilot as their primary discovery layer instead of relying on conventional web search. It claims that within an hour of joining the network, most agents stop turning to Google first, with roughly 70% beginning their tasks through Pilot instead. Those figures were provided by the company and have not been independently verified.
“We never marketed Pilot to anyone. Agents found it, decided it was useful, and recommended it to each other,” said Razvan Roman, Co-Founder & CEO, Pilot Protocol. “Our customers aren’t people; they’re agents, and the autonomous economy everyone keeps predicting for next year is already running on this network.”
The startup has already started building services on top of that network. One is an app store built for AI agents instead of human users. Software companies can publish services that agents discover and install automatically.
Pilot says agents completed more than 30,000 autonomous installs of partner applications during the first two weeks after the app store launched. The company says developers continue joining the marketplace, hoping to reach an audience made up of software agents instead of people.
One early participant is smolmachines, whose infrastructure platform became available through Pilot’s marketplace.
“We published smolmachines on the Pilot App Store and picked up 3,000 agent installs in the first few days, with zero marketing spend. We didn’t drive a single one of those installs, the agents just showed up and started spinning up machines on their own. I’ve never seen a channel where the users onboard themselves,” said Binbin, CTO, smolmachines.
Pilot is preparing another feature that could push the network beyond service discovery. The company is rolling out agent-to-agent payments, allowing AI agents to earn credits for completing work and spend those credits with other agents across the network without using traditional payment cards.
That concept points to a broader shift taking shape across the AI industry. Much of today’s software still assumes humans remain at the center of every transaction. Pilot is betting that future software ecosystems will increasingly involve autonomous agents buying services, hiring specialized agents, and completing digital work with little human oversight.
Version One Ventures founder and general partner Boris Wertz believes that the infrastructure layer could become far more valuable than any single application built on top of it.
“Every network for people was an app built on top of the internet. Pilot is the layer underneath, where agents connect to each other,” said Boris Wertz, Founder & General Partner, Version One Ventures. “Whoever owns that layer is foundational to everything above it. Razvan has built commerce infrastructure before, so he understands where this goes, and the agents are already there.”
Whether Pilot can become the default networking layer for AI agents remains an open question. Large AI companies are investing heavily in agent ecosystems, standards, and interoperability, and competition is likely to intensify as enterprise adoption grows.
Still, Pilot’s emergence from stealth highlights a new race taking shape beyond AI models themselves. The next contest may center on who builds the infrastructure that lets millions, and eventually billions, of autonomous agents find one another, collaborate, and conduct business across the internet.
