Agent identity, attestations and pay-per-call are all live on this chain: deployed, and barely inhabited. We measured how much of that is actually used, and paid nineteen sellers from our own wallet to find out who delivers.
Measured 2026-09-09 by Prooflines, from public RPC and explorer data. Every number below can be recomputed by anyone; the method is at the bottom of the page. Not affiliated with Monad Foundation.
The short version: nothing here is missing. The contracts are deployed, the rails settle, and a payment goes through in under a second. What is missing is people using them, and a way to tell a seller that works from one that takes the money anyway.
Our discovery index held 1,608 resources advertising a Monad rail, up from 1,090 two weeks earlier. We asked all 1,607 reachable ones for their offer the way a buying agent would. 1,533 returned a live Monad rail.
Those 1,533 endpoints collapse to nine domains and twenty-one payout addresses. Two sellers account for 87% of them: theaslangroupllc.com with 849 and agent402.tools with 546. Everyone except us settles in USDC; we are still the only seller taking native MON. Two former sellers, orth.sh and one vercel.app deployment, now have no live Monad rail at all.
We bought one cheapest callable endpoint from each payout address we could afford: nineteen purchases, 0.404 USDC, every one reconciled against the chain rather than trusted.
| Outcome | Calls | What it looks like |
|---|---|---|
| delivered | 12 | HTTP 200 with real data: technology lookups, liquidity for a pair on Base, token balances for an address. |
| paid, nothing back | 7 | Payment settled on chain, then HTTP 400, 401, 404 or 500, including one seller failing its own published example. |
37% of what we paid for produced nothing. A separate problem sits inside the twelve that did deliver: one returned HTTP 200 with real data and no settlement header, so our own accounting recorded less spent than the chain did. A client that trusts the settlement header instead of reconciling on chain under-counts its own spending. We saw the same seller do the same thing two weeks earlier, which is why the reconciliation runs against USDC transfer logs rather than against what the sellers report.
This is the part no directory shows, and it only appears if you pay more than once per seller.
orthogonal.com settles into sixteen addresses. We paid fifteen of them. Eight delivered and seven took the money and returned an error, with four different status codes, so it is not one broken service but an absence of any check that a paid call can be served at all. Read by domain, this seller looks like a working business with over a hundred endpoints. Read by address, it works half the time. Elsewhere the same logic runs the other way: glim.sh and cascade.fyi are two brands paying into one address, so they are one seller.
The official identity and reputation registries are deployed on Monad mainnet and hold real volume. The volume is not what it looks like.
10,251 agents are registered. Registrations by month: 542 in June, 1,250 in July, then 46 in August and 18 in September. The totals are historical waves, and the current organic rate is around fifteen a week. There are 9,187 feedback events from 7,689 distinct clients, but they land on only 83 agents. So 0.8% of registered agents have ever been rated, each rating client rates about 1.2 times, and the rated few carry roughly a hundred ratings each. That shape, many single-use raters and few targets, is what reputation farming looks like from the outside, and it is why a reputation number here should not be read as a reputation.
The Ethereum Attestation Service is the primitive several recommended designs assume. It is on Monad. It is also empty.
| Network | Contract | Schemas | Attestations |
|---|---|---|---|
| mainnet | 0xcea6ea33…88eb (EAS v1.4.0) | — | 5 |
| mainnet | 0x7184adab…8242 (registry v1.4.0) 0xf1850692…9654 (registry v1.3.0) |
14 | — |
| testnet | 0x321e37e2…eaf8 (registry v1.3.0) | 20 | 0 |
Five attestations in the whole history of mainnet, from one attester, on one schema. None at all on testnet, where anyone building would start. None of these sit at EAS's canonical addresses, which is why a check by address alone says EAS is absent: our first pass concluded exactly that, and was wrong. Sweeping by event topic found it.
Discovery from our public x402 index; one unpaid request per endpoint, using the seller's own declared method. Purchases from a single wallet, one per payout address, reconciled against USDC Transfer logs rather than against the sellers' own responses. ERC-8004 and EAS counts come from event logs read from genesis (Attested, Registered, and the registry events our tracker already publishes). Anyone can repeat all of it with a public RPC and an explorer API.
What this does not say: that these standards do not work. They do. A payment settles, an attestation records, a registration lands. Nor does a low number prove a bad design; an empty registry on a chain this young may simply be early. And our own gate is in this data too, because leaving the mapmaker off the map would be its own kind of dishonesty. Figures move; the feeds behind them are live.