A2X

Concepts

Agent monetization

Why A2X exists

Agents today can talk — but they cannot transact. Every existing agent framework assumes a trusted caller on the same wallet, inside the same company, behind the same auth. That forces every monetization attempt into a bilateral contract: API keys, invoices, SaaS subscriptions, rate plans. None of that composes.

A2X is built around a different bet: an agent should be a public endpoint that charges per call, on-chain, to any anonymous caller. The building blocks for that bet already exist in open specs:

  • A2A — an open JSON-RPC protocol for agent-to-agent calls.
  • x402 — HTTP 402 with a machine-readable payment envelope, settled on-chain.

A2X is the TypeScript layer that glues them together and makes the loop easy:

  1. Expose an agent over A2A with one SDK call.
  2. Gate the A2A endpoint with x402.
  3. Earn per call. No accounts, no invoices.

What "monetization" means here

Three concrete properties, in order of importance:

Permissionless. Any caller — human, agent, or script — can pay and call. No onboarding, no keys.

Per-call. Pricing is scoped to the call, not to the relationship. A long-running session is many billed calls.

Settled outside the app. Payment verification is cryptographic; settlement is on-chain. The agent never holds funds in escrow, never reconciles invoices, never runs a billing backend.

What it is not

  • Not a marketplace. A2X doesn't take a cut; it doesn't even know what your agent costs.
  • Not a wallet. You bring the wallet; A2X verifies the payment header.
  • Not a gateway. Your agent runs where you want it to run (Fly, Vercel, your laptop).

Roadmap shape

Layer Today Next
A2A server Shipping (toA2x, A2XAgent) More transport bindings
A2A client Shipping (A2XClient) Payment-aware retry loop
x402 server middleware Manual wiring (see Quickstart) First-class SDK helpers
x402 client retry Manual wiring Automatic 402 → pay → retry
Agent discovery & pricing AgentCard-embedded Indexed catalog

Next