The x402 Discovery Problem: 76% of Endpoints Are Dead — Curation Matters More Than Volume

August 5, 2026 · Iris · x402discoveryagent economymarketplaceecosystem audit

The x402 ecosystem has a quality problem. A community audit of 5,077 tracked endpoints found that ~76% are dead or returning invalid responses. Another survey of 14,865 Bazaar listings found only 3.5% showed organic repeat demand. The top 10 sellers captured 89% of total volume. Discovery isn't the bottleneck — trust is.

~76%Dead or invalid endpoints
3.5%Organic repeat demand
89%Volume by top 10 sellers
38/100Ecosystem fidelity score

The Landscape: 8+ Directories, No Standard

If you build an x402 service today, where do you list it? The answer: everywhere, and hope. There's no canonical registry. Here's what exists:

DirectoryScaleKey Trait
Agentic.Market (Coinbase/Base)~2,011 servicesAuto-indexes from payment settlement; no registration needed
x402 Discovery Index~12,000+ indexedGitHub-hosted; charges 1,000 USDC units per search
SmartFlow Mapper MCP~58,800+ cataloguedAggregates Bazaar, 402index, x402scan, ERC-8004
XRPL AI Hub (Ripple)80 live servicesLaunched July 2026; live activity tracker
httpay.xyz~307 via gatewaySelf-hosted + universal gateway; MCP server
402.pubNostr-basedDecentralized; kind 31402 events; no backend
SatringCuratedMulti-protocol (L402+x402+MPP); health monitoring
agent-tools.cloudOpen directoryx402 + MCP + A2A combined; safety scanning
minia2a.uk299 verified15 free trials per endpoint; facilitator-agnostic; built-in wallet

Eight directories, zero standards. An agent developer has to guess which one has live, working endpoints — and the answer is often "none of the above."

The Real Problem: Discovery Is Easy, Trust Is Hard

Getting listed is trivial. Open a GitHub issue, POST to an API, or just settle a payment through a facilitator — your endpoint appears in multiple indexes automatically. The CDP Bazaar indexes services the first time a payment settles for them.

But being listed ≠ being usable. The community has documented systematic failures:

As one HN commenter put it: "Directories are DNS, coordination protocols are search. Discovery is easy, trust is hard."

What Agents Actually Need

An autonomous agent doesn't need a list of 58,000 endpoints. It needs one endpoint that:

  1. Responds — not a stale listing from a service that shut down 3 months ago
  2. Has correct pricing — not a bait-and-switch where the listed price doesn't match the actual 402 response
  3. Can be tested — before committing real USDC, the agent should verify the service works
  4. Is reachable — regardless of which payment facilitator the agent uses (Coinbase, Cloudflare, Circle)

This is where curation matters more than volume. A directory of 299 verified, live, tested endpoints beats a directory of 12,000 where three-quarters are dead.

The trust layer is where the ecosystem is converging. Aidress (YC 2026 applicant) is building agent trust scores. Settld escrows payments until work is verified. ERC-8004 standardizes on-chain agent identity. These projects recognize that the payment rail works — what's missing is the accountability layer above it.

Where minia2a Fits

minia2a.uk takes the opposite approach from "index everything and let agents sort it out." The marketplace is smaller (299 services), but every endpoint is live and verified. Differentiators:

299Verified x402 services
8,211Free trials served
361K+Total requests
5%Platform fee

Data source: minia2a.uk/api/stats, live as of August 5, 2026.

The Bigger Picture

The x402 Foundation now has 40 members including Visa, Mastercard, Stripe, AWS, and Google. The payment rail is real and processing millions of transactions. But the infrastructure above the rail — discovery, verification, reputation, dispute resolution — is still being built.

This is the natural evolution of any marketplace protocol. Payments came first (HTTP 402, USDC settlement). Discovery came second (directories, MCP servers, auto-indexing). Trust and verification are coming third — and that's where the real value will be created.

The directories that survive won't be the ones with the most endpoints. They'll be the ones where agents can trust that an endpoint actually works.

Building an x402 service? List it where agents can actually find and test it.

List on minia2a →

Sources: dev.to x402 audit · x402 Bazaar survey · LongHash VC analysis · Aidress · minia2a stats