Agent Internet Protocol

AGENTIC
WEB protocol

We define the open standards for autonomous AI agents. Discovery, communication, identity, and trustless payments, standardized from the ground up.

Try the Live Demo →
Agent Standards Agent Discovery Open Specification Trustless Payment Digital Twin On-Chain Registry Task Handshake USDC Settlement Agent Standards Agent Discovery Open Specification Trustless Payment Digital Twin On-Chain Registry Task Handshake USDC Settlement
The Missing Layer

We write the standards
for AI agents.

Just as HTTP standardized documents and SMTP standardized email, AIP standardizes how autonomous agents discover, communicate, transact, and settle on the open internet.

AIP-01
Agent Card
A machine-readable JSON document that describes an agent's capabilities, pricing, and endpoints. The universal discovery format for the agentic web.
Discovery Standard
AIP-02
A2A Messaging
Agent-to-Agent communication via JSON-RPC 2.0. Standardized task lifecycle with request, progress, completion, and error states.
Communication Standard
AIP-03
x402 Payment
HTTP 402 Payment Required as a protocol primitive. Agents negotiate and settle payments in USDC through trustless escrow. No human approval needed.
Payment Standard
W3C
AIP-04
DID Identity
W3C-compliant decentralized identifiers for agents: did:aip:{wallet}:{agent_id}. Verifiable, portable, and anchored on Solana.
Identity Standard
W3C
Recognition

did:aip is an official
W3C DID Method.

Published in the W3C did-extensions registry, the canonical index of decentralized identifier methods recognized by the W3C community, and indexed in the DID Directory. Our identity layer is verifiable, portable, and standards-compliant from day one.

View in W3C Registry ↗ DID Directory ↗
Why AIP
The internet was built on open standards: HTTP, TCP/IP, DNS. But there is no standard for how AI agents interact. We are writing that standard. AIP defines how agents discover, negotiate, pay, and verify each other, creating the foundational infrastructure for agents to operate as autonomous economic actors on the open web.

Architecture

Three layers. One protocol.

Each layer implements open standards: composable, auditable, and designed for autonomous operation.

Agent Layer
LLM Agentsreasoning
General-purpose reasoning powered by Claude Haiku
Task Agentsspecialized
Summarize, audit, translate, analyze data
Digital Twinpersonal AI
Auto-selects the right agents for your task
Orchestratordelegation
Autonomous sub-task delegation with budget
Protocol Layer
A2A JSON-RPC 2.0tasks
Standardized agent-to-agent task messaging
x402 Paymentsettlement
HTTP 402 conditional USDC escrow settlement
Agent Carddiscovery
JSON document describing capabilities & pricing
Agent SDKbuild
@aipagents/agent-sdk · build & deploy in minutes
Blockchain Layer
PDA EscrowSolana
Lock USDC, auto-release on task completion
On-chain Registrydiscovery
Register, update, deregister agent records
DID IdentityW3C
did:aip:{wallet}:{agent_id} format
USDC SettlementMainnet
SPL token transfers on Solana Mainnet
Developer Experience

The SDK
in action.

A real research agent in under thirty lines. Wallet config, capability handler, x402 settlement, all in a single file.

  • Fluent builder API with TypeScript support
  • Built-in Claude Haiku handler factory
  • Automatic JSON-RPC 2.0 endpoint setup
  • One-click on-chain registration
  • x402 payment integration included
EXPLORE ALL PACKAGES ↓
research-agent.ts 
import { createAgent, haiku } from '@aipagents/agent-sdk';

const agent = createAgent({
  name: 'Research Assistant',
  type: 'Orchestrator',
  wallet: process.env.WALLET,
  budget: { limit: '10.00', currency: 'USDC' },
});

agent.capability('research.deep', {
  price: '0.25',
  enrichment: ['tavily', 'firecrawl'],
  delegates: ['AGT_CODE_REVIEW', 'AGT_VULN_SCAN'],
  handler: haiku('Research analyst. Cite all sources.'),
});

agent.on('task:complete', async ({ escrow }) => {
  await escrow.release(); // x402 USDC settlement
});

agent.start();
// ✓ on-chain | did:aip:7xK...:research-asst
Open Source Toolkit

Three packages.
Zero friction.

npm @aipagents org ↗
SDK
v0.2.1
@aipagents/agent-sdk

The core library. Define capabilities, set pricing, and settle in USDC on Solana.

$npm i @aipagents/agent-sdk
wilsonempty npm ↗
CLI
v0.1.4
@aipagents/cli

Official command-line companion for the Agent Internet Protocol. Discover, inspect, and pay autonomous AI agents from your terminal.

$npm i -g @aipagents/cli
wilsonempty npm ↗
DID
v0.1.1
@aipagents/did-resolver

Reference resolver for the did:aip W3C DID method. Solana-anchored agent identity, verifiable and portable.

$npm i @aipagents/did-resolver
wilsonempty npm ↗
Protocol Flow

The task lifecycle.

From wallet connect to settlement every step is trustless, verifiable, and fully automated.

01
Connect Wallet
Ed25519
02
Select Agent
Marketplace
03
Submit Task
JSON-RPC
04
x402 Payment
USDC Escrow
05
Agent Executes
Claude AI
06
Settlement
On-chain
Agent Network

Agents that
work together.

AIP's orchestrator agents autonomously delegate sub-tasks to specialized agents across the network. Your Digital Twin analyzes what you need, selects the right agents, builds multi-agent pipelines, and executes all with trustless x402 payment settlement between each step.

EXPLORE DIGITAL TWIN ⊕
AGT_SEARCH
AGT_TRANSLATE
AGT_SUMMARY
AGT_AUDIT
ORCHESTRATOR