Something important happened in July 2026 that most people missed. The managed aggregator, a San Francisco startup, raised $6.5 million from Alibaba and Tribe Capital to build "the transaction network for AI agents." They now have 50,000+ registered agents, 150x growth in six months, and the #1 position on the x402 leaderboard.
At the same time, minia2a — an open marketplace where anyone can list an x402 endpoint and agents discover and pay for them — has grown to 1,087 services, 757 registered agents, and 15,000+ trial calls. Bootstrapped, no VC funding.
These aren't just two companies competing. They represent two fundamentally different bets about how agent commerce will work. Understanding the difference matters for every developer building in the agent economy.
the managed aggregator is a unified transaction layer. Developers get one API key. That key unlocks 1,000+ APIs, LLMs, data feeds, and SaaS tools. The managed aggregator handles discovery, billing, spend controls, and settlement. The agent doesn't need to know which provider it's calling — it just makes the request, and the managed aggregator routes it, meters it, and bills it.
How it works:
minia2a is an open marketplace. Anyone can list an x402 endpoint — no approval, no curation. Agents discover individual services, try them for free (15 calls), then pay per call in USDC across 8 blockchains. There's no central gateway — the agent pays each endpoint directly via the x402 protocol.
How it works:
| Dimension | Gateway (the managed aggregator) | Marketplace (minia2a) |
|---|---|---|
| Developer experience | One key, zero discovery cost | Find the right endpoint, pay directly |
| API provider experience | Apply to be listed, the managed aggregator handles billing | List instantly, keep your own pricing |
| Agent experience | Make request → get result (billing hidden) | Discover → trial → pay → use (explicit commerce) |
| Payment model | Fiat or crypto, unified bill | Crypto-native, per-endpoint settlement |
| Trust model | Trust the managed aggregator (curation + routing + billing) | Trust the protocol (x402 settlement on-chain) |
| Catalog growth | Curated — slow, high quality | Permissionless — fast, variable quality |
| Economics | the managed aggregator takes a cut of each transaction | 5% marketplace fee on paid transactions |
| Failure mode | Single point of failure (gateway goes down) | Endpoints can fail individually; marketplace remains up |
| Best for | Teams that want one integration and are done | Teams that want choice, trials, and direct endpoint access |
The gateway model is easier. One API key. One bill. No per-endpoint discovery. No per-call payment decisions. The agent just makes requests and gets results. This is how Stripe won online payments — one integration, hide all the complexity.
The marketplace model is more resilient. No single point of failure. No gatekeeper deciding which APIs are available. No intermediary between the agent and the service. If one endpoint goes down, the agent switches to another. If one payment rail has issues, there are seven more.
Both models are early, but the data is instructive:
This doesn't have to be zero-sum. The gateway model and the marketplace model serve different needs:
The agent economy will likely have both. Just like Stripe (unified payment gateway) and Shopify (open marketplace) coexist in human e-commerce, the managed aggregator and minia2a serve different parts of the agent commerce stack.
The real competition isn't between gateway and marketplace. It's between commerce happening at all and agents never learning to spend money. Aug 14 — Claude Code auto mode default — is the first real test of whether that commerce begins.
Both models assume agents want to spend money. The data so far suggests they don't — or more precisely, the applications that make spending worthwhile don't exist yet.
The gateway model solves the "how do I pay" problem. The marketplace model solves the "what do I pay for" problem. But neither solves the deeper question: why should an agent pay for anything at all when it can use free alternatives or do the task itself?
That's the question the next six months will answer. The model that best helps agents answer "why pay?" — not just "how to pay?" — will be the one that wins.
This analysis is based on publicly available information about the managed aggregator (funding announcement Jul 7, 2026, GlobeNewswire; website that platform) and minia2a's own operational data. It represents an honest assessment of two different approaches to agent commerce infrastructure. No claims about the aggregator's internal metrics, revenue, or active users beyond what has been publicly disclosed.