Spend-control tools limit how much an agent spends. Bellwether checks whether the thing being bought is any good, using real paid test calls rather than self-reported listings.
Built for developers running agents that buy data and API calls, x402 API sellers who want a credibility signal, and funds testing agent commerce.
| Endpoint | Price | p50 | Verdict |
|---|---|---|---|
| api.tidewatch.dev/v1/quote | 0.004 USDC | 412 ms | caution |
| feeds.marlin.xyz/pools | 0.002 USDC | 188 ms | allow |
| sigscan.io/x402/wallet-risk | 0.010 USDC | 950 ms | allow |
| news.harborline.ai/summary | 0.015 USDC | 2.4 s | avoid |
| geo.pinpoint.run/resolve | 0.001 USDC | 143 ms | allow |
A latency spike between probes 13 and 15 coincided with two
Paste an x402 URL and Bellwether pays for a few test calls from its own probe wallet. It then scores schema match, latency, price history and the rate at which paid calls fail to deliver.
Planned| Check | Result | Weight | Status |
|---|---|---|---|
| Schema match | 14 of 14 fields | 35% | pass |
| Latency | p50 188 ms, p95 420 ms | 20% | pass |
| Price history | 0.002 USDC for 21 days | 15% | pass |
| Post-payment failures | 0 of 31 paid calls | 30% | pass |
| Overall | Score 94 of 100 | allow |
Registry endpoints are probed on a schedule, every 30 minutes for most. Each probe records whether payment settled, whether the body matched the last known schema, and how long it took.
PlannedTwo days dipped below 47 because of short Solana RPC retries
Agents call the score endpoint with the target URL and get allow, caution or avoid plus reasons they can log or act on. The call itself costs nothing.
Planned
curl "https://api.bellwether.dev/v1/score?url=https://api.tidewatch.dev/v1/quote" \
-H "Authorization: Bearer bw_live_..."
{
"verdict": "caution",
"score": 71,
"price": { "amount": "0.004", "asset": "USDC", "changed_7d": false },
"reasons": [
"asOf field returned as string in last 3 probes",
"1 of 24 paid calls returned an empty body"
],
"last_probed": "2 min ago"
}Every registry endpoint gets a public report page with its current verdict, uptime timeline and price history. Sellers can link it from their docs as proof of quality that they cannot edit.
PlannedWatch an endpoint and choose what matters: price moves, schema drift, rising failure rates or downtime. Alerts go to email or a webhook your agent can read.
PlannedSpend-control tools limit how much an agent spends. Bellwether checks whether the thing being bought is any good, using real paid test calls rather than self-reported listings.
Every screen reads from the same model. Live parts read public onchain data; nothing ever asks your wallet to sign.
Read the docs{
"product": "Bellwether",
"network": "solana",
"features": [
{
"name": "Paste an x402 endpoint URL and get a scorecard: ",
"status": "planned"
},
{
"name": "Scheduled probes on a small registry of Solana-s",
"status": "planned"
},
{
"name": "Free JSON score API an agent can call before pay",
"status": "planned"
},
{
"name": "Shareable endpoint report pages that sellers can",
"status": "planned"
}
]
}Live features work today. Planned ones are next on the roadmap, in this order.
Short answers. The docs have the rest.
A probe-and-score service that tells an agent whether an x402 paid endpoint is worth paying for before it pays. Spend-control tools limit how much an agent spends. Bellwether checks whether the thing being bought is any good, using real paid test calls rather than self-reported listings.
Bellwether is in development. Every feature is on the status list with what ships next, and the app opens in a sample workspace so you can see how it will work.
No. You can explore without one. Connecting a Solana wallet only shares your public address so Bellwether can read your balances; it never asks you to sign a transaction.
Developers running agents that buy data and API calls, x402 API sellers who want a credibility signal, and funds testing agent commerce.
A probe-and-score service that tells an agent whether an x402 paid endpoint is worth paying for before it pays.