Skip to main content
A-Identity gives every AI agent two things it does not have today: a way to prove who it is, and a way to pay. Identity comes first (Know Your Agent), then settlement in stablecoins, under permissions that both sides agree to. It is built on three open protocols. We did not invent them. We connect them into one agent-native stack. The one rule: verify first, then pay.
Five stages in a row: AI agent and owner, verify identity and KYA, bound with the policy engine, pay in USDC on Arc, settle with a receipt and audit trail

The same loop as a picture: the owner sets limits, identity and KYA are checked on-chain, the policy engine bounds the spend, USDC settles, and a receipt lands in the audit trail. The board was drawn for Arc testnet, the phase-1 network. The loop now also runs on the mainnet rails listed under Chains.

ERC-8004

Portable on-chain identity and reputation for agents. A-Identity reads live ERC-8004 registries deployed on Circle Arc testnet.

x402

Pay-per-action over the HTTP 402 standard, settled in stablecoins.

MCP

The Model Context Protocol. How agents reach your tools, and pay for them.

​
What You Can Build

Verify An Agent

Resolve an agent’s identity and reputation before you transact with it.

Bound What It Can Spend

Set a limit that’s enforced three ways, including an on-chain revert on Arc.

Charge Per Action

Wrap any tool or endpoint in an x402 paywall and get paid in USDC, on-chain or gasless and sub-cent with Nanopayments.

Connect Over MCP

Expose your service to agents through a Model Context Protocol server.

Embed The SDK

Drop identity and payments into your own agent in a few lines.

​
A Note On Safety

Anything that holds a key, deploys a contract, or moves real value stays human-on-the-loop. A person approves it. Agents act inside the limits and policies you set. The idea is a human in the tower, not in the driver’s seat.