The economy will be run by agents.
Software already writes, plans, and codes. Soon it will buy, sell, negotiate, and hire. A growing share of economic activity will pass through agents acting for people and companies.
For most of history, markets forced a choice. A local shop knows its customers but cannot reach far. A website reaches everyone and treats everyone the same. Agents are the first systems that can do both: search every option, hold full context, and settle on terms shaped for one buyer, at the scale of the whole market.
That only works if agents can reach each other. Today they cannot.
An agent cannot reliably find another agent across companies, tools, and marketplaces. It cannot carry context from one interaction to the next without being handed everything again. It cannot prove who it is, or verify who it is dealing with. It cannot agree to terms and move money without a person approving each step.
The web was built for people clicking pages. APIs were built for one program talking to one other under a single contract. Neither was built for millions of agents that have to discover each other, trust each other, share context, negotiate, and transact across systems no single party controls.
That layer does not exist. We are here to build it.
AgentNet is an open project for the connective layer of the agent economy: discovery, identity, trust, context, agent-to-agent protocols, standards, and safe commerce between machines. We work in the open because an economy shared by everyone cannot run on a stack owned by one company.
Three things we hold to
- 01Do whatever moves the agent economy closer, sooner.
- 02The gains it creates should reach everyone, not a few.
- 03Open first. Ideas, protocols, and research, shared as we go.
The shift is coming regardless. We would rather it arrive open, safe, and built to serve the people it runs for.