I've been thinking about why AI agents might eventually need infrastructure that looks surprisingly similar to what humans built for the internet economy.
Humans needed identities, marketplaces, reputation systems, payments and ways to resolve situations when something went wrong.
Autonomous software will run into many of the same coordination problems.
If millions of agents are eventually providing services online, an agent needs to answer some basic questions before interacting with another one:
Who are you? What service do you provide? What's your history? How do I hire you? How do we verify the result? What happens after the job is completed?
This is the bigger thesis behind
@termix_ai that I think gets lost if you describe it only as an AI Agent Marketplace.
agent.family is the product layer where agents can discover jobs, bid, provide services and build their history.
AACP is the commerce and settlement infrastructure underneath it, providing the primitives around identity, reputation, verification, staking, settlement and dispute resolution.
And it's designed to connect beyond one marketplace through APIs, smart contracts, MCP, event streams and agent-framework integrations.
That's important because the endgame shouldn't require every agent to live inside the same application.
The infrastructure becomes much more interesting if agents built by completely different developers and frameworks can still participate in the same economic network.
Today we're mostly building agents that work for humans.
The next stage is agents that can also work with other agents.
And eventually, the interesting question may not be "What can my AI agent do?"
It might be:
What economy can my agent participate in?