Two platforms. Same problem. Radically different answers. Satring and minia2a both help AI agents find paid APIs. But they represent two competing philosophies about how discovery should work: one is a comprehensive directory listing everything, the other is a trial-first marketplace where agents test before they buy. The difference isn't just UX — it's a bet on how the agent economy will actually grow.
Satring is an open-source, protocol-agnostic directory of paid agent APIs. It crawls, indexes, and health-checks APIs across three protocols: L402 (129 services), x402 (672 services), and MPP (407 services). Listings include pricing in satoshis, 30-day hit counts, category tags, and whether the domain is verified.
How it works:
The philosophy: List everything. Let the market sort out quality. More services = more utility. The directory is the product.
minia2a is a curated marketplace where every service has free trials. Agents don't browse a catalog and hope the API is good — they call it, see the response, and decide. Trial data (what agents actually use) becomes the discovery signal.
How it works:
The philosophy: Curate. Let agents try before they commit. Usage data, not listing count, is the product.
| Dimension | Satring | minia2a |
|---|---|---|
| Services listed | 845 | 299 |
| Protocols supported | 3 (L402, x402, MPP) | 1 (x402) |
| Payment currency | Bitcoin / Lightning (sats) | USDC (Base, multi-chain) |
| Free trials | No | Yes — 7,440 used |
| Real usage data | 30d hit counts | Per-endpoint trial counts, users, retention |
| Health monitoring | 95.6% health, live probing | Implicit (trial failures surface issues) |
| Domain verification | 28 verified | Curated onboarding |
| Community ratings | 29 ratings | None (trial data is the signal) |
| Open source | Yes | No |
| MCP integration | No | Yes — 300 tools via HTTP/SSE |
| Agents with wallets | N/A (no wallet system) | 42 |
| Credits purchased | N/A | 2,984 ($29.84) |
Satring's 845 services span 9 categories. The protocol breakdown — 672 x402, 407 MPP, 129 L402 — reflects where developer activity is: x402 dominates, MPP is growing fast (Stripe-backed), and L402 has the Lightning early-adopter base. The 95.6% health rate tells us these are real, maintained endpoints, not abandoned listings.
But here's what the directory model doesn't tell us: are agents actually using any of these APIs? 30-day hit counts give a clue — the top-listed services have 66–92 hits/month — but "hits" don't distinguish between a human browsing and an agent transacting.
minia2a's 7,440 trials paint a different picture than any directory could. The top 10 endpoints capture 53.5% of all trials. The #1 endpoint (CAPTCHA solving, 902 trials, 133 users) isn't something a directory would rank highly — it's a commodity utility. The real discovery signal is trial velocity: what agents try repeatedly, across many users, over time.
This reveals something the directory model can't: what agents need isn't the same as what developers build. The most-listed categories on directories (finance, AI/ML tools) don't match the most-used categories in practice (utilities, memory, web infrastructure).
Satring's domain verification (28 of 845 services) proves the submitter controls the domain. That's useful — it prevents impersonation. But it doesn't prove the API works well for agents.
minia2a's trust signal is different: 7,440 trials, 318 unique users, actual responses. When 133 different agents try the same CAPTCHA solver, that's a trust signal no star rating can replicate. The API might not have a verified domain, but agents are voting with their trial credits.
Neither model has solved trust completely. Satring has 29 ratings across 845 services (3.4% rated). minia2a has trial data but no way for agents to flag broken endpoints or report bad responses. Both are missing the layer that matters most: delivery verification — proving the work was actually done, not just paid for.
This is the most consequential difference, and it's bigger than either platform:
This isn't just a currency choice — it's a bet on which ecosystem agent developers will prefer. Bitcoin/Lightning has lower fees per transaction but requires agents to manage BTC exposure and Lightning channels. USDC/Base has stable value and simpler wallet infrastructure but higher gas costs and a 5% facilitator fee.
The market hasn't decided. Satring's protocol mix (L402+x402, L402+MPP) suggests developers are hedging — listing on multiple protocols simultaneously. This multi-protocol reality is probably where the market is heading.
The ideal agent API discovery platform probably combines elements of both models:
Neither Satring nor minia2a has all five. The platform that gets there first wins the discovery layer — and as we've argued before, when settlement gets commoditized, discovery wins.
The agent API discovery layer is still wide open. Two credible platforms with two credible philosophies are competing to define how agents find services. The directory model says "list everything, let the market sort." The trial-first model says "let agents test, let data decide." Both have real traction. Neither has won.
But one thing is already clear from 7,440 trials: agents don't browse directories — they try endpoints. The discovery platform that makes trying the easiest wins.