EXPERIMENTALOpen Source Community

Agent Network Protocol (ANP)

2 min read

Decentralized protocol for trustless agent discovery and collaboration.

SpecificationGitHubDocs

Overview

How can billions of AI agents from different organizations discover and collaborate without centralized intermediaries?

ANP answers this with a fully decentralized approach, using W3C Decentralized Identifiers (DIDs) for identity and semantic web technologies for capability description.

Three-Layer Architecture

Layer 1: Identity & Encryption

  • Decentralized Identifiers (DIDs): Self-sovereign agent identity
  • End-to-End Encryption: DID-based key exchange
  • Verifiable Credentials: Proof of capabilities/permissions

Layer 2: Meta-Protocol

  • Protocol Discovery: Agents advertise supported protocols
  • Capability Negotiation: Dynamic agreement on interaction
  • Self-Organization: No central coordination needed

Layer 3: Application

  • JSON-LD: Semantic capability descriptions
  • RDF: Rich relationship modeling
  • Flexible Protocols: Application-specific extensions

Key Differentiator: Decentralization

Feature A2A ANP
Discovery Agent Cards (centralized) DIDs (decentralized)
Identity OAuth/API Keys Decentralized Identifiers
Trust Platform-based Cryptographic
Governance Linux Foundation Open community

Security & Privacy

ANP's DID-based architecture provides:

  • Cryptographic Signing: All messages verifiable
  • Selective Disclosure: Reveal only what's needed
  • Trustless Operation: No trusted intermediaries

When to Use ANP

Use ANP for:

  • Decentralized agent marketplaces
  • Cross-organizational collaboration without authority
  • Privacy-preserving interactions
  • Open agent ecosystems

Consider A2A for:

  • Enterprise deployments needing mature tooling
  • Centralized discovery is acceptable
Example Code
# ANP Agent Discovery (Conceptual)
from anp import ANPAgent, DIDResolver

# Create agent with DID identity
agent = ANPAgent(
    did="did:anp:agent123",
    capabilities=["translation", "summarization"]
)

# Discover agents by capability
resolver = DIDResolver()
translators = await resolver.discover_by_capability(
    capability="translation",
    languages=["en", "es"]
)

# Connect and invoke
for translator in translators:
    session = await agent.connect(translator.did)
    result = await session.invoke("translate", {
        "text": "Hello, world!",
        "target": "es"
    })
Quick Info
Architecturedecentralized
Message FormatJSON-LD
TransportHTTP/HTTPS, WebSocket, DIDComm
SDKsPython, TypeScript
Agent-to-Agent Features
Discovery
Negotiation
Streaming
Bidirectional
Multi-Agent
Ready to Evaluate?
Agent Playground
Test against real-world scenarios
RepKit
Reputation SDK for agents
Consulting
We'll implement it for you
Stay Updated

Get notified when protocols evolve.

Subscribe
Related Resources
Evaluation Patterns Failure Modes
Strengths
  • Fully decentralized—no central authority
  • Strong privacy through DID-based identity
  • Semantic interoperability
  • W3C standards compliance
  • Future-proof for open ecosystems
Considerations
  • Early stage (experimental)
  • Smaller ecosystem
  • Complex implementation
  • Requires DID infrastructure
Tags
decentralizeddidsemantic-webprivacy
Back to Protocols

Compare with Other Protocols

Compare against:(2/2)
Agent Network Protocol (ANP)(current)
Strengths
Agent Network Protocol (ANP)
experimental
  • Fully decentralized—no central authority
  • Strong privacy through DID-based identity
  • Semantic interoperability
AG-UI Protocol
beta
  • Purpose-built for frontend integration
  • Human-in-the-loop patterns built-in
  • Framework-agnostic design
Agent Name Service (ANS)
experimental
  • Leverages existing DNS infrastructure
  • Cryptographic identity via PKI
  • Scalable and decentralized
Considerations
Agent Network Protocol (ANP)
  • • Early stage (experimental)
  • • Smaller ecosystem
  • • Complex implementation
AG-UI Protocol
  • • Frontend-specific scope
  • • No agent discovery
  • • Smaller community than MCP/A2A
Agent Name Service (ANS)
  • • Early draft stage
  • • Requires PKI infrastructure
  • • DNS propagation delays
Best For
Agent Network Protocol (ANP)
Decentralized agent marketplacesCross-organizational collaborationPrivacy-preserving networksFederated research systems
AG-UI Protocol
AI-powered chat interfacesCollaborative document editingApproval workflow UIsReal-time agent dashboards
Agent Name Service (ANS)
Finding agents by capabilityVerifying agent identityCross-platform discovery
Technical Details
Agent Network Protocol (ANP)AG-UI ProtocolAgent Name Service (ANS)
Architecturedecentralizedevent-drivendistributed
Message FormatJSON-LDJSON EventsDNS Records
TransportHTTP/HTTPS, WebSocket, DIDCommSSE, WebSocket, HTTP WebhooksDNS
Version0.11.0draft-00
SDKsPython, TypeScriptTypeScript, Python, React—
Architecture
Agent Network Protocol (ANP)decentralized
AG-UI Protocolevent-driven
Agent Name Service (ANS)distributed
Message Format
Agent Network Protocol (ANP)JSON-LD
AG-UI ProtocolJSON Events
Agent Name Service (ANS)DNS Records
Transport
Agent Network Protocol (ANP)HTTP/HTTPS, WebSocket, DIDComm
AG-UI ProtocolSSE, WebSocket, HTTP Webhooks
Agent Name Service (ANS)DNS
Version
Agent Network Protocol (ANP)0.1
AG-UI Protocol1.0
Agent Name Service (ANS)draft-00
SDKs
Agent Network Protocol (ANP)Python, TypeScript
AG-UI ProtocolTypeScript, Python, React
Agent Name Service (ANS)—

All Protocols & Initiatives

9 standards documented
AG-UI Protocol

The standard for streaming AI agent interactions to user interfaces.

Supporters: CopilotKit, Microsoft, Google...
Agent Name Service (ANS)

DNS-based discovery system for AI agents with PKI trust.

Supporters: IETF Contributors
Agent Network Protocol (ANP)(current)

Decentralized protocol for trustless agent discovery and collaboration.

Supporters: ANP Community, Academic Research
Agent Protocol

A standardized REST API specification for exposing AI agents as services.

Supporters: AI Engineer Foundation, AutoGPT, W3C Community Group
Agent-to-Agent Protocol (A2A)

The standard for AI agents to communicate, collaborate, and delegate tasks.

Supporters: Google, Microsoft, IBM...
Model Context Protocol (MCP)

The universal standard for connecting AI agents to external tools and data.

Supporters: Anthropic, OpenAI, Google...
NANDA

Framework for decentralized AI agent coordination and lifecycle management.

Supporters: MIT, Academic Research
Open Agent Specification (Agent Spec)

Declarative language for portable agent definitions with built-in evaluation.

Supporters: Oracle, CopilotKit
Open Agentic Schema Framework (OASF)

Standardized schemas for defining AI agent capabilities and metadata.

Supporters: AGNTCY, Cisco (Outshift)