The agent payment infrastructure is being built at astonishing speed. In the last 72 hours alone: Stripe acquired OpenRouter for $10B, Cloudflare launched agent wallets with USDC accounts, and GoPlausible processed $111K across 160K x402 transactions on Algorand with a 100% success rate.
But all of this is payment rails. The question nobody's asking: once agents have wallets, where do they go to spend the money?
That's the discovery layer. And it's splitting into two competing models.
Satring is an open-source, multi-protocol API directory indexing 845 services across L402, x402, and MPP. It treats API metadata as a paid product: you get 5 free lookups per day per IP address, after which every search, listing, or detail view costs money.
How it works:
The philosophy: discovery has value, so charge for it. Every API call to the directory itself is a microtransaction. If you're an agent that needs to find 10 APIs to solve a task, you're paying Satring $0.50+ before you've made a single useful call.
minia2a takes the opposite approach: discovery is free, and every listed API includes 15 free trials. Agents don't pay to find services — they pay to use them at scale. The platform takes a 5% fee on transactions.
How it works:
The philosophy: let agents prove value before committing capital. Discovery is the top of the funnel — charging for it is like charging people to walk into a store.
| Dimension | Satring (Paid Directory) | minia2a (Trial-First Marketplace) |
|---|---|---|
| Services listed | 845 | 298 |
| Verified/curated | 28 (3.3%) | 298 (100% — every one has a working endpoint) |
| Protocols | L402 + x402 + MPP | x402 |
| Free tier | 5 requests/day, thin summary only | 15 free calls per endpoint, full access |
| Cost to discover | $0.05/request after free tier | $0 — always free |
| Cost to list | $0.50 | $0 |
| Cost to rate | $0.01 | Free (usage-based quality signals) |
| Payment rail | Bitcoin/Lightning | USDC (Base) |
| Paying users | Unknown (no public data) | 42 wallets, 2,984 credits purchased |
| Real usage data | Health probes only | Per-endpoint trial counts, retention metrics, user counts |
| GitHub | 0 stars, 2 commits | Active development |
| API docs | OpenAPI/Swagger | REST + MCP (HTTP) |
| Agent UX | Agents hit a paywall to learn what an API does | Agents learn by calling the API itself, not reading metadata |
Agents don't browse directories like humans browse app stores. An agent deciding whether to call x402-captcha-solve doesn't want to read a description — it wants to call the endpoint and see if the response solves its problem.
This is the fundamental insight that the trial-first model gets right: for AI agents, the API call IS the evaluation. A 200 OK with a solved CAPTCHA is worth a thousand words of documentation. A paywall that returns "here's what this service is called but not how to use it" is worse than useless — it's an obstacle.
The data supports this. On minia2a, the most-trialed endpoint is CAPTCHA solving (841 trials, 133 users) because agents encounter CAPTCHAs in the wild and need to solve them now. They don't browse a directory to "discover" CAPTCHA solving — they search for it when blocked. A free trial lets them verify it works in 200ms and move on. A $0.05 metadata lookup followed by uncertainty about whether the API actually delivers is a non-starter.
Satring's model has internal logic. If discovery is valuable, charge for it. The $0.01 rating fee is a clever anti-spam mechanism — it costs money to leave fake reviews, so they're more likely to be genuine. The multi-protocol approach (L402 + x402 + MPP) means agents using any payment rail can find compatible services.
But the model has a critical flaw at this stage of market development: agents don't have wallets yet. The entire M2M payment ecosystem is pre-scale. Asking an agent to pay $0.05 to read about an API before it has even decided whether to use it is putting the tollbooth at the wrong end of the highway.
Neither model has fully solved discovery. Satring has breadth (845 services, 3 protocols) but no depth (28 verified, thin free tier). minia2a has depth (100% curated, 15 free trials, real usage data) but narrower protocol coverage (x402 only).
The winning platform will combine:
If you're building an API for AI agents, you have a choice about where to list it:
List on a paid directory (Satring): Pay $0.50 to submit, hope agents discover you among 845 services, and accept that they'll see a "thin summary" of your API unless they pay. Your API competes on metadata quality.
List on a trial-first marketplace (minia2a): List for free, give agents 15 free calls to prove your API works, and earn when they convert to paid usage. Your API competes on actual performance — response quality, latency, reliability.
In a market where agents make decisions based on API responses, not marketing copy, the second option isn't just cheaper — it's structurally better aligned with how AI agents actually evaluate tools.
The discovery layer is the last unbundled piece of the agent payment stack. Wallets exist (Coinbase, Cloudflare). Settlement exists (x402, MPP, L402). But nobody has won discovery yet.
Satring's paid-directory model is internally consistent but misaligned with current market reality: agents don't have wallets, charging for metadata reduces discovery, and 3.3% verification rate signals a quantity-over-quality approach.
The trial-first model — free trials, free listing, monetize usage — is better positioned for the pre-scale phase of the M2M economy. The platform that helps agents prove value will capture the market that emerges when agents have value to spend.
The window is open. The models are competing. Agents will decide.
minia2a is a trial-first marketplace for x402 agent APIs. 298 services, 15 free calls each, USDC on Base. minia2a.uk