Permissionless AI agents mainnet
Praxis Protocol launched its mainnet to enable permissionless AI agent networks that can discover, coordinate and transact without central platforms, signalling a push toward decentralised agent ecosystems. The announcement positions open agent‑to‑agent marketplaces as an alternative architecture for coordination and monetisation in the agent economy (x.com).
Most artificial intelligence agents still work like apps in a mall: they sit inside one company’s building, use that company’s payment rails, and cannot easily hire tools from outside. Praxis says its mainnet is live with a different model, where agents can find each other directly and transact on a peer-to-peer network instead of through a central platform. (docs.prxs.ai) (prxs.ai) Praxis describes that network as a mesh, which is the internet version of a neighborhood street grid instead of a single highway toll booth. Its documentation says agents discover peers automatically through peer-to-peer networking, then coordinate through agent-to-agent protocols built for task delegation and distributed workflows. (docs.prxs.ai 1) (docs.prxs.ai 2) The basic promise is simple: one agent can ask another agent for help without a human opening five tabs and wiring the pieces together by hand. Praxis says agents on the network can share tools through the Model Context Protocol, execute workflows in natural language or a domain-specific language, and collaborate on tasks through intelligent delegation. (docs.prxs.ai) Finding an agent is the first bottleneck in any open network, so Praxis built a directory layer on top of blockchain records. Its Explorer indexes agents that follow the ERC-8004 identity standard and lets users search them by name, domain, capability, and blockchain address across Ethereum-compatible networks. (docs.prxs.ai) Paying an agent is the second bottleneck, because a marketplace is not very open if every transaction still depends on a platform operator. Praxis says agents can earn, spend, and trade with the PRXS token, and its 2026 roadmap says the “Genesis Block” milestone is the first real United States dollar coin earned by a non-team provider on mainnet. (docs.prxs.ai) (prxs.ai) That puts Praxis in the same lane as a wider push to give agents their own internet plumbing instead of forcing them through software built for humans clicking buttons. A recent Agent Network Protocol white paper from the World Wide Web Consortium community group argues that today’s internet creates data silos and high collaboration costs for agents, and calls for standardized protocols for interconnection. (w3c-cg.github.io) Praxis is also leaning hard on the idea that open networks become more useful as more participants join. Its public roadmap lists mainnet deployment, a distributed registry for decentralized discovery, 10 or more first-party reference agents, a provider dashboard for earnings and reputation, and a later goal of 1,000 active agents. (prxs.ai) The hard part starts after launch, not at launch. Praxis’s own materials promise reputation scores, identity verification, signature validation, staking, and security audits, which is another way of saying an open agent economy still needs ways to stop spam, fraud, and unreliable bots from flooding the market. (docs.prxs.ai) (prxs.ai) If this model works, using an artificial intelligence agent starts to look less like installing one giant app and more like hiring a tiny contractor that can subcontract other specialists on its own. Praxis’s bet is that discovery, payment, identity, and coordination can all be moved into open network rails, so the next agent marketplace looks more like email or the web than a closed software store. (docs.prxs.ai 1) (docs.prxs.ai 2)
