Everyone's building an agent marketplace. That's what the headlines say. But today I did something simple: I curled every agent API marketplace I could find, live, at 21:45 UTC on August 4, 2026.
Two of them are dead. One of them doesn't resolve DNS. Three are front pages with no visible agent product. One is real competition.
Here's the scorecard — and what it tells us about the state of the M2M micropayment market.
| Marketplace | Status | Model | Last GitHub Activity |
|---|---|---|---|
| minia2a.uk | ✓ Live | Platform directory + free trials | Active (today) |
| Satring | ✓ Live | Curated API directory | N/A (closed source) |
| MoltMart | ✗ 404 | P2P x402 marketplace | March 28, 2026 |
| HYRVE AI | ✗ DNS fail | Agent community marketplace | N/A |
| agenton.me | ? Front only | Unknown | N/A |
| taskbounty.com | ? Front only | Task bounty board | N/A |
| aisa.one | ? Front only | Agent platform | N/A |
MoltMart was the most interesting competitor: a peer-to-peer x402 marketplace on Base, with ERC-8004 on-chain identities ($0.05 mint), direct seller-to-buyer payments, no escrow. The architecture was solid — Next.js 15 frontend, FastAPI backend, PostgreSQL on Railway, x402 middleware for payment verification. 240 commits across three contributors (two AI agents, one human).
But here's what happened when I checked:
$ curl -s -o /dev/null -w "HTTP %{http_code}" https://moltmart.app
HTTP 404
$ curl -s -o /dev/null -w "HTTP %{http_code}" https://api.moltmart.app/health
HTTP 404
Both the website and API return 404 — "Application not found." The GitHub repo hasn't been updated since March 28, 2026. Four months of silence. The deployed application appears to have been taken down or never launched beyond testnet.
This isn't a judgment on the team — they built something real, open-source, with a clean architecture. But it's a data point: building a marketplace is harder than building a directory. MoltMart tackled P2P payments, identity minting, contract deployment — the full stack. That's a lot of infrastructure to keep running.
HYRVE claimed 5,750+ community members, 85% commission rates, and CashClaw middleware. As of today:
$ curl -s -o /dev/null -w "HTTP %{http_code}" https://hyrve.ai
HTTP 000 # DNS resolution failure
The domain doesn't resolve. Whether this is a temporary outage or a permanent shutdown is unclear, but for agents trying to discover APIs right now, it's a dead end.
These three return HTTP 200 — their landing pages load. But none of them have an obvious agent API marketplace product. They might be building in stealth, or they might be general AI platforms that got lumped into the "agent marketplace" category by over-eager competitive researchers. Without a public API or agent-facing product, they're not competitors yet.
Satring (satring.com) is the only other live, functional agent API directory I can find. Here's the comparison:
| Metric | Satring | minia2a |
|---|---|---|
| Services listed | 845 | 297 |
| Verified services | 28 (3.3%) | All (platform-hosted) |
| Protocols | L402, x402, MPP | x402 |
| Free trials | 10/day for search | 15/endpoint, every endpoint |
| Model | Crawler + curation | Platform + hosting |
| Health monitoring | Yes | Yes (built-in) |
| Hosting | Unknown | AWS Ireland (EU) |
Satring's crawler approach gives them a larger raw listing count (845 vs 297), but only 3.3% of their listings are verified. The other 96.7% may or may not work. minia2a hosts every service directly, so every endpoint is verified by definition — it either responds or it doesn't.
Satring's multi-protocol support (L402 + x402 + MPP) is genuinely interesting and gives agents more payment options. This is something worth watching.
Looking at the pattern, three reasons emerge:
API providers won't list without agents. Agents won't browse without APIs. MoltMart solved the supply side (P2P — anyone can list) but couldn't generate enough demand to justify keeping the infrastructure running. A marketplace with 0 buyers is just an expensive directory.
With Cloudflare Wallets launching today (August 4) and the x402 Foundation backed by Visa, Mastercard, Stripe, and Coinbase, the payment layer is becoming a commodity. The differentiator isn't "we support x402" anymore — it's what you build on top of it. MoltMart had x402. They still died.
APIs consume compute. Health monitoring costs bandwidth. Payment verification costs gas (even if it's fractions of a cent). Free trials multiply these costs. If you're not generating revenue — and most early-stage marketplaces aren't — you're burning someone's money. At some point, that stops.
The agent marketplace layer is consolidating faster than anyone expected. Six months ago there were half a dozen projects claiming to be "the app store for AI agents." Today there are two — and one of them is a crawler, not a platform.
This doesn't mean the market is dead. It means it's real. Real markets have winners and losers. The x402 protocol is seeing 160M+ transactions, $41-50M in cumulative volume, and Cloudflare just entered with institutional-grade wallet infrastructure. The agents are coming. The question is whether the discovery layer — the layer where agents find out what to buy — can survive long enough to serve them.
What do the two survivors (Satring and minia2a) have in common?
The agent marketplace graveyard is filling up — and that's actually good news for the ecosystem. It means real competition is happening. The projects that survive will be the ones that solve the chicken-and-egg problem without burning out, keep infrastructure costs manageable, and offer something the payment rails don't: discovery with trust.
The x402 protocol tells an agent how to pay. What it doesn't tell them is what's worth paying for. That's the marketplace layer. And as of August 4, 2026, there are only two places agents can go to answer that question.
Data collected live on August 4, 2026 at 21:45 UTC. All competitor URLs were curled from an independent server. minia2a stats: 297 services, 6,351 trials, 318 agent users, 42 wallets, 347,755 total requests — from /api/stats.