Infrastructure for the Machine Economy

The machine economy is not coming. It's here. But almost nobody can see it yet, because the numbers everyone looks at are wrong.

We run the largest x402 facilitator by daily transaction volume. We overtook Coinbase in December 2025. Since then, we've settled over 25 million payments across seven blockchain networks. We verify the quality of 5,000+ API endpoints every fifteen minutes using real payments. We publish reputation scores to on-chain registries on both Base and Solana.

We see the data that most people analyzing this space don't have access to.

What's real

The x402 protocol works. An agent hits an API, gets a 402 response with payment terms, signs a USDC authorization, retries with a payment header, and receives the data. Settlement happens in seconds. No API keys, no accounts, no invoices. The payment IS the authentication.

This is a genuine architectural breakthrough. For the first time, any piece of software on the internet can charge for access without requiring a business relationship with the requester. A solo developer in Lagos can build an API, protect it with x402, and have agents paying for it within hours.

The supply side is growing fast. Six months ago, fewer than 100 merchants accepted x402 payments. Today there are over 4,000 qualified sellers — services with real transactions from multiple unique buyers.

What's noise

Here's what we know from operating the infrastructure: the vast majority of x402 transaction volume is not organic commerce. It's leaderboard gaming.

When x402scan launched as the ecosystem explorer, it created a visible ranking system. Facilitators, API sellers, and even blockchain networks began running circular funding loops to inflate their position on the leaderboards. One chain passed a protocol-level update to provide $1M in free gas specifically for x402 settlements.

This isn't unique to x402. Every marketplace in history has dealt with fake volume. But it means the headline numbers — "$100M in transactions," "100M+ payments" — don't represent real agent commerce. The honest number, after filtering wash trading, is roughly $1.6 million in real volume across the entire ecosystem over the last 30 days.

That's early. We know it's early. But it's real, and it's growing.

What matters now

Two independent research reports — OnchainLu's "Machine Economy 2030" and Dynamic's "Agentic Finance and Stablecoins" — arrived at the same conclusion: discovery is the largest unsolved problem in the agentic commerce stack.

When an agent needs a lead enrichment API, a satellite imagery provider, or a flight booking service, how does it find one? Today, the answer is hardcoded tool lists and curated marketplaces. That doesn't scale to millions of agents and millions of services.

Whoever builds the ranking and recommendation system that routes agents to the right service — reliably, with quality guarantees — captures enormous leverage. That system needs three things that nobody else has built together: settlement infrastructure (to see real payment data), quality verification (to test endpoints with actual money), and on-chain reputation (to make quality signals portable and verifiable).

What we're building

Dexter spans the three layers where value concentrates in the agentic commerce stack.

Settlement. Our facilitator processes more daily x402 volume than any other, including Coinbase's. Every settlement generates data — which endpoints deliver, which fail, which are worth paying for. This data is the foundation of everything else.

Quality. Every fifteen minutes, our verification system probes endpoints across the marketplace. It makes real x402 payments, evaluates response quality using frontier AI models, and assigns composite scores from 0 to 100. This isn't metadata scraping. We're paying real money to test real APIs and publishing the results.

Trust. We're one of the most active publishers of ERC-8004 reputation data — the emerging on-chain standard for agent identity and trust. Every settlement through our facilitator generates on-chain feedback. Every verification cycle publishes quality assessments to IPFS with cryptographic hash verification on Base and Solana. This creates a data moat that compounds with every transaction.

The market we see

By 2030, the machine economy could represent $800 billion to $1.4 trillion in annual agent-initiated commerce. That estimate comes from straightforward math: one billion knowledge workers, each with agents spending $3-5 per day on data, tools, and services that currently cost orders of magnitude more through traditional channels.

We're not there yet. But the infrastructure being built now will determine who captures value when that transition happens. The protocols are live. The supply side is growing. The trust layer is being standardized. And the teams building today are writing the defaults that billions of agents will eventually run on.

We intend to be one of those defaults.