OKX AI Agent Marketplace Hits 20,000 Tasks as Machines Pay Machines

OKX AI has spent the summer building something unusual: a marketplace where artificial intelligence agents hire, pay and review other AI agents for narrowly defined digital jobs, often for less than the price of a cup of coffee. The OKX AI agent marketplace went live on July 1, 2026, and in the weeks since, it has quietly become a working test case for what happens when machines start paying machines directly, without a human clicking “approve” on every transaction.

Key takeaways

• The OKX AI agent marketplace launched on July 1, 2026, letting AI agents discover, hire and pay other agents for specialized tasks.
• More than 10,000 agent identities have registered, alongside over 4,000 provider listings.
• On August 10, more than 2,700 agents were active and over 20,000 tasks had been created.
• Software services lead sales at 36%, followed by art creation at 34%, with lifestyle and finance each at 14%.
• OKX will host an in-person developer hackathon in Singapore on October 6, 2026, to accelerate new agent-powered services.

Launch and Growth of OKX AI Agent Marketplace

The core idea behind the marketplace is straightforward: developers supply specific capabilities, and other agents or people buy access to exactly the task they need, nothing more. Rather than one large AI system trying to do everything, the model breaks digital work into small, purchasable units that any agent can plug into on demand.

That structure points toward what OKX describes as a future of one-person companies, where a single entrepreneur directs a network of AI agents instead of hiring a full team. Those agents, in turn, could purchase services from other agents — running code, scanning markets, or checking a crypto token for security risks — rather than building every capability in-house.

The growth numbers suggest the concept found an audience fast. Since the July 1 launch, OKX AI has registered more than 10,000 agent identities and approved over 4,000 provider listings. By August 10, more than 2,700 agents were active on a single day, and the platform had logged over 20,000 tasks created in total. Those figures describe a marketplace that moved from concept to functioning economy within about six weeks.

Service Categories and Sales Distribution

Software services and art creation dominate the OKX AI agent marketplace, together accounting for the large majority of recorded transactions. That split shows demand skewing toward practical automation and creative generation over other categories.

• Software services: 36% of recorded sales
• Art creation: 34% of recorded sales
• Lifestyle services: about 14%
• Finance services: about 14%

Beyond those four buckets, the catalog already includes agents built for narrower purposes — retrieving news and social-media intelligence, pulling structured crypto market data, or running blockchain security checks. Listings such as Newsliquid, which profiles X users and tracks influencer audiences, and CoinMW, which returns public data from platforms including X, Instagram, TikTok, Reddit and YouTube, illustrate how granular the pay-per-call model has become. CoinAnk OpenAPI, a crypto derivatives data engine offering roughly 80 market-data services covering price charts, ETF flows, funding rates and whale activity, has reportedly logged more than 14,900 sales on its own, an early sign of how demand concentrates around a handful of high-utility listings.

Notable Marketplace Agents and Pricing Models

Three listings stand out as early examples of how agent-to-agent AI services get priced and consumed inside the marketplace: a compute sandbox, a security-scanning tool, and a crypto trading signal provider.

TinyDock’s sandboxed compute service

With TinyDock, an agent can execute a brief Python or Node script inside a disposable, network-isolated Firecracker microVM — a short-lived virtual machine that has no internet connectivity and is destroyed as soon as the program completes. The calling agent pays 0.01 USDT, roughly one U.S. penny, and receives the output without needing an account, an API key, or any human involvement. Each sandbox is destroyed once the program exits, making TinyDock a clean example of machine-to-machine commerce: one agent buying a narrowly defined computing task from another, at a price too small to justify a traditional invoice.

CertiK’s crypto token security scans

CertiK brings established blockchain-security capabilities into the pay-per-call model, offering token scans, Skynet Scores that summarize a project’s risk profile, transaction traces, balance and state-change lookups, and source-code checks. Each call for these crypto security scans costs just 0.001 USDT. The pricing shows how a full security audit can be broken into small, on-demand functions that other agents purchase and fold directly into their own workflows, rather than commissioning a standalone review.

1M’s crypto trading signals

1M supplies AI crypto trading signals covering major tokens including BTC, ETH and SOL. A one-off report costs 1 USDT and gives a read on possible price movement along with suggested trade parameters. Buyers who want continuous coverage can instead pay 20 USDT a month for a subscription that monitors the market around the clock. Subscribers can act on the signals manually or connect them to OKX’s Agent Trade Kit to automatically open, close or adjust positions — a detail that shows how quickly these micro-services can plug into fully automated trading loops.

Upcoming Developer Hackathon in Singapore

OKX plans to bring developers together in person on October 6, 2026, for a hackathon in Singapore aimed at turning new ideas for agent-powered services into working products. The event is designed to accelerate the early momentum already visible in the marketplace’s registration and sales figures, giving builders a concentrated window to launch listings that could join the ranks of TinyDock, CertiK and 1M.

Why this matters for the wider industry is fairly direct: if agents can reliably discover, hire and pay each other for fractions of a cent, the cost of assembling complex digital workflows drops sharply, and the barrier to running a lean, largely automated business falls with it. The open question is whether that same low-friction model can scale without running into the quality-control, verification and trust issues that come with agents transacting autonomously, at volume, with little human oversight in the loop.

FAQ

What is the OKX AI agent-to-agent marketplace?

It is a digital economy platform where AI agents discover, hire, and pay other agents for specialized services.

When was the OKX AI marketplace launched?

The marketplace was launched on July 1, 2026.

What types of services are most commonly sold on the marketplace?

Software services and art creation lead sales, accounting for 36% and 34% respectively, followed by lifestyle and finance services.

What are some examples of service pricing in the marketplace?

TinyDock charges 0.01 USDT per short program run, CertiK charges 0.001 USDT per security scan call, and 1M offers crypto trading signals starting at 1 USDT.

Article produced with the assistance of artificial intelligence and reviewed by the editorial team.