Agent Network ProtocolAn open-source protocol enabling secure, decentralized communication among AI agents.

The Agent Network Protocol (ANP) is an open-source protocol enabling secure, decentralized communication among AI agents. It is a layered protocol stack that reuses the open Internet foundation and is designed for open, cross-domain agent networks. The protocol has three layers: the bottom layer reuses HTTP, CA, DNS, CDN, Search, TLS, and the existing web operating model; the middle layer provides decentralized identity (DID:WBA), handles, authentication, key material, secure messaging, and cross-domain communication foundations; the top layer includes Agent Description, Agent Discovery, and domain-specific protocols for payment, authorization, authentication, commerce, and other agent workflows. ANP allows agents to publish capabilities, find interoperable services, and enable direct chat, groups, encryption, attachments, and federation semantics. The protocol is organized in modular sections such as identity and communication, description and discovery, messaging profiles, and domain protocols. It features DID:WBA identity, WNS handles, authentication, key distribution, and secure messaging. This allows for a verifiable agent identity and readable handles, and enables cross-domain trust roots. ANP can be adopted by implementations as a foundation for their application-specific needs and allows for the addition of messaging, discovery, and payment capabilities as needed. The goal of ANP is to create a decentralized infrastructure for AI agents. The protocol allows for new specifications and updates, with recent additions including new WBA Name Space, end-to-end instant messaging, and an agent payment protocol. It follows a three-layer architecture with the open Internet infrastructure at the base, identity and communication infrastructure above it, and application protocols at the top. ANP's modular design allows for adoption and extension by multiple parties and enables open, cross-domain agent networks.