OKX, the crypto exchange valued at $25 billion after NYSE parent Intercontinental Exchange invested $200 million in March, is opening a marketplace where AI agents hire one another, settle payments autonomously in stablecoins, and build portable on-chain reputations. Called OKX AI, the platform opens to developers on July 1 after a closed beta with 50 early AI service providers, according to TechCrunch.
The launch is the first major attempt by a crypto exchange to build dedicated financial infrastructure for agent-to-agent commerce. While most agent builders have focused on what agents can do, OKX is betting on how they pay for it.
The Problem OKX Is Solving
AI agents are proliferating fast. A November 2025 MIT Sloan/BCG report found 35% of surveyed businesses had already deployed agents, with another 44% planning implementation soon. Gartner projects AI agent software spending will reach $206.5 billion in 2026, rising to $376.3 billion by 2027, according to ZDNET.
But deployment is outrunning infrastructure. When one agent needs data from another agent, or when a coding agent needs a security audit agent to check a smart contract before execution, there is no standardized way for them to discover each other, negotiate terms, settle payments, and verify that the work was done. Traditional payment rails were not designed for micropayments between autonomous software. A $0.003 query to a market data service does not fit Visa’s fee structure.
“Traditional financial infrastructure was built for humans. The agentic economy needs infrastructure designed for autonomous software,” Star Xu, founder and CEO of OKX, told TechCrunch. “The coming decade will be defined by one-person companies that generate over a million dollars in annual revenue, because every individual effectively gains an unlimited workforce.”
How OKX AI Works
The marketplace builds on OKX’s Onchain OS toolkit, which previously enabled AI agents to hold digital wallets, make stablecoin payments, and establish persistent identities on-chain. OKX AI extends this into a four-layer stack:
Discovery. Agents register services on the marketplace with defined capabilities, pricing, and performance history. Other agents can search for and evaluate providers based on on-chain reputation scores built from completed transactions.
Payments. Transactions settle in stablecoins on blockchain rails, enabling micropayments that would be impractical through conventional payment processors. An agent querying live market data at $0.001 per request can settle hundreds of transactions per minute without batching delays or minimum transaction thresholds.
Identity. Each agent maintains a persistent on-chain identity tied to its transaction history, allowing reputation to be portable across platforms rather than locked to a single marketplace.
Dispute resolution. GenLayer, a launch partner, provides what co-founder and CEO Albert Castellana described as “essentially a digital court system” for agents to resolve contractual disagreements. “The challenge for us is distribution. OKX already has that,” Castellana told TechCrunch.
Developers access the platform through Onchain OS. No OKX account is required to get started, and the platform is compatible with Claude Code, Codex, Hermes, and OpenClaw.
The Launch Partners
Three initial service providers illustrate the marketplace’s scope:
CertiK offers on-chain security assessments, letting AI agents evaluate the safety of a crypto wallet or token before executing a transaction. This is runtime security verification at machine speed, not a manual audit process.
CoinAnk provides live market data on a pay-per-query basis. Instead of subscribing to a data feed, agents pay only for the queries they run, settling each transaction individually in stablecoins.
GenLayer handles dispute resolution. When an agent claims a service was not delivered as specified, GenLayer’s infrastructure arbitrates the disagreement on-chain.
OKX is applying its existing fraud detection, compliance systems, and exchange infrastructure to the marketplace. Haider Rafique, OKX’s chief marketing officer and global managing partner, told TechCrunch that the rollout will proceed in phases before becoming more widely available.
The Competitive Landscape for Agent Commerce
OKX is not building in a vacuum. The infrastructure for autonomous agent transactions is becoming a contested space.
On June 29, a coalition including Google, IBM, Circle, UiPath, Stellar, Cardano, and Hedera launched the Legal Context Protocol (LCP), an open-source Apache 2.0 framework enabling AI agents to transact on behalf of humans with verifiable legal terms, consent, and dispute resolution. LCP approaches the problem from the legal compliance side rather than the payments side.
China’s State Administration for Market Regulation (SAMR) released a national standard for AI agent identity management on the same day, establishing a unified framework for all AI agents operating within the country.
OKX’s advantage is that it already has 150 million users and a functioning payment infrastructure. Its disadvantage is that the marketplace launches into a crypto-native ecosystem, which limits its reach to developers already comfortable with blockchain tooling. Rafique argued that OKX’s existing network of crypto developers will seed the marketplace, while its broader strategy “extends well beyond digital assets.”
The company’s ambitions are tied to two parallel plays. The ICE partnership is about “modernizing markets” through tokenization. OKX AI is about “modernizing money” for autonomous software. Both bets require OKX to become something larger than a crypto exchange.
India as the Developer Funnel
OKX is targeting India as a primary developer market, even though the company suspended its crypto trading services in the country in 2024 while navigating regulatory requirements. Rafique told TechCrunch that India remains a high-priority market and that developer products like OKX AI face fewer regulatory hurdles than spot crypto trading.
This is a notable strategy: using agent infrastructure tooling as a re-entry point into a market where your core business is restricted.
The Trillion-Dollar Projection
Rafique said OKX believes “agentic commerce” could become a trillion-dollar market over the next five years. That projection is aggressive. But the underlying logic follows a clear trajectory: if agents are handling an increasing share of operational tasks for businesses, and those agents need to procure data, compute, security audits, and other services from each other, then the volume of agent-to-agent transactions will grow proportionally to the number of deployed agents.
The critical question is whether those transactions will route through a crypto-native marketplace or through conventional payment infrastructure that adapts to handle micropayments. Stripe has been experimenting with sub-cent transaction processing. PayPal has micropayment APIs. Neither was designed for autonomous software, but both have larger merchant networks than OKX.
OKX’s bet is that blockchain-native payments will win on three dimensions that conventional rails cannot match: settlement speed (instant vs. batched), transaction minimums (no floor vs. minimum fees), and programmability (smart contracts vs. static APIs). Whether those advantages matter enough to overcome the friction of requiring stablecoins remains an open question.
The Architectural Risk
The deeper risk is structural. A marketplace where agents transact autonomously needs guardrails that do not yet exist. What happens when an agent hires another agent that produces incorrect output? GenLayer’s dispute resolution handles explicit contract violations, but detecting subtle quality failures in machine-generated work requires evaluation capabilities that current systems lack.
There is also the question of cascading spend. An agent authorized to spend $50 on market research could, in theory, hire 500 agents at $0.10 each without triggering any spending alert designed for human transaction patterns. OKX’s fraud detection systems were built for human traders. Adapting them for autonomous agent spending patterns is a non-trivial engineering problem.
These are solvable problems. But they are the problems that will determine whether agent commerce infrastructure becomes a real market or a demo-stage curiosity.