Payments for the Agent Economy

Software is starting to act on its own. Agents read, decide, write code, and coordinate with other systems. The one thing they still cannot do well is move money.

We think payments are the next piece of infrastructure the agent economy needs, and that it has to be built deliberately, not bolted onto wallets designed for people. That belief is why Q402 exists.

Why we started

We kept hitting the same contradiction: a wallet can hold USDC and still be unable to pay. Native gas, chain-specific execution, approval flows, receipts, and the question of what an autonomous system should even be allowed to do all sit in the gap between intent and settlement.

A person can pause, buy gas, switch networks, approve, retry, and save a receipt by hand. For a human that gap is friction. For an agent it is a wall. Q402 started as an attempt to remove the wall, the way payments infrastructure has always matured: by pushing the hard parts down into a layer the application no longer has to think about.

What we believe

Four convictions shape everything we build.

This is already real

Q402 is shipped, not a whitepaper. The rail settles real stablecoin payments today across 11 EVM networks, in USDC and USDT, with the payer never holding a single gas token. Any agent can call it over MCP from Claude, Codex, Cursor, or Cline, the same way it calls any other tool.

Under the surface there is more: dedicated Agent Wallets with spending caps, policy Hooks that run before money moves, recurring payments, a cross-chain gas bridge, and a verifiable receipt for every settlement. But the point of this note is not the feature list. The point is the direction, and the size of what is coming.

The market is already moving

Two curves are about to cross. Stablecoins have quietly become one of the largest settlement networks on earth: by several industry trackers they moved tens of trillions of dollars onchain over the past year, on the order of the major card networks, and the volume keeps compounding. At the same time, AI agents are leaving the demo stage and entering production, and analysts expect the agentic software market to grow into the hundreds of billions of dollars this decade.

Put those together. Software that can act, meeting money that already moves at the speed of software. Once an agent can decide, negotiate, and execute on its own, the one missing primitive is payment, and almost every agent that does anything economic will eventually need to pay or be paid. That is not a niche. It is the default surface area of the next internet.

Where we are going

From here we are building toward agent-to-agent commerce, where one agent invoices and another settles with no human in the loop; verifiable reputation, so agents can choose who to trust by an onchain track record instead of a confident sentence; and self-driving treasuries that keep working capital productive and in the right place, always inside policy their owners control.

The direction is simple to state: the user defines the rules, the agent expresses the intent, and Q402 turns approved intent into settlement.

We are building for the day when hundreds of millions of people have agents acting for them, and most payments begin with software rather than a human hand. We intend to be the settlement layer beneath all of it, the quiet rail under a billion small decisions a day.