Agentic Commerce on The OP Stack
Roughly 71 percent of stablecoin transfer volume is already driven by automated systems. Not people, bots, scripts, AI agents executing payments, rebalancing positions, and settling transactions without anyone approving individual steps. That share is growing as enterprises deploy agentic capabilities at scale.
Most are deploying on infrastructure they don't own. Which means at scale, products deploying on other chains will run out of blockspace. on your own chain, you control scale and throughput for your agents.
The chains capturing agentic order flow today didn't get there by accident. They made a specific infrastructure decision early, and the transaction volume followed.
The pattern
OKX built OnchainOS for X Layer, its OP Stack chain. OnchainOS includes APP — Agent Payments Protocol — a standard for full agentic commerce: quoting, negotiation, escrow, metering, settlement. Complete autonomous business relationships between agents, handled end to end. OKX has 120 million users and ICE backing at a $25 billion valuation. APP on X Layer isn't an experiment. It's production infrastructure for a business that needed to own its agent transaction economics.
Celo migrated its entire network to OP Stack to serve the same category from a different angle. Sub-cent transaction costs, fee abstraction that lets agents pay gas in USDC or USDT with no native token required, no separate treasury to manage.
Circle nanopayments (transactions settling fractions of a cent, autonomously, without human approval) only work when the gas cost doesn't consume the payment. Celo's fee abstraction makes that math work at the infrastructure level.
The pattern across both: own the chain, build the agent toolkit on top of it, capture the revenue that flows through it.
What your chain needs to run this
High-frequency agent transactions require reserved blockspace. Sequencer policies that keep agent order flow from competing with general traffic when it matters. Custom gas rules (sponsored transactions, zero-gas models) so enterprises can absorb costs on behalf of their agent workflows without building workarounds. These decisions only exist when you own the sequencer.
The authorization layer is what's been missing longest. ERC-7715 is now live on Optimism via MetaMask, supported across 12 chains. It gives agents scoped permissions to transact autonomously within defined limits, without full account control. The agent acts inside of a pre-defined scope.
OP Stack includes sequencer policy controls, custom gas rules, ERC-7715 compatibility, and throughput designed for high-frequency traffic. OKX and Celo didn't build this infrastructure from scratch. They started from OP Stack and built their specific agent commerce capabilities on top of it.
The revenue question
A top-3 US exchange generated $75 million in sequencer revenue in the second half of 2025. That's what owning the chain your users transact on is worth, at current volumes, before agentic commerce adds a new transaction category on top. Messari's analysis of the same period: $1 in transaction fees generated $10 in application revenue across OP Stack chains, $415 million across all partners in six months.
Agentic commerce will generate order flow at a scale human-driven transactions don't approach. Agents handling payments, subscriptions, and settlements continuously, at high frequency, without manual approval loops. Every enterprise deploying these capabilities on infrastructure they don't own is routing that order flow (and the economics attached to it) somewhere else.
What moving looks like
Enterprises making chain infrastructure decisions now are 8 to 12 weeks from decision to production. OKX and Celo are already running agentic commerce on OP Stack. The infrastructure is MIT licensed, every line is public, and if a better option exists for your workload, you should be able to find it.
To learn more visit: https://www.optimism.io/