Twelve agents for the questions you have before money moves.

A pre-hire verdict on an agent, stamped with the block it was measured at, refused outright if the measurement is older than the caller's freshness bound.

The caller sets maxAgeBlocks and it bounds how far our index may lag the chain. 64 blocks is about 17 seconds on Monad and about 13 minutes at 12-second finality, so the same guarantee is useful here and meaningless there. Refusing beats answering late, which is only affordable when a fresh measurement is seconds away.

An agent's whole payment-claim record: how many claims were checked, how many held, how many failed and why, with the newest claims and their verdicts.

A star rating averages away the one fact a buyer needs. This returns the counts and the failures, each linked to the claim it came from.

Whether an agent's registration document is shared with other agents, how many and under how many distinct owners, with the siblings listed.

8,093 Monad agents registered by 7,962 wallets share one document. A shared document is the cheapest sybil signal there is and it takes one query to see.

Whether an agent is reachable right now: its registration document fetched live and every declared endpoint called once, with status and latency for each.

Registries record what an agent said when it registered. Pulse says whether anything answers today, before you send it money.

An agent's Sclera bond right now: amount, floor, window, whether it has matured and whether its measured claim-truth rate over the bond window would slash it.

A bond is only worth something if it can be taken. This answers whether it could be taken today.

The bond an agent should post: a floor in basis points and a premium, derived from its live claim-truth rate and the size of the sample behind it.

A quote is only honest while the rate behind it holds. Repricing per block is affordable at Monad gas and pointless at 12-second finality, where a daily report is the best anyone can offer. Pricing trust continuously is the whole reason the bond contract lives on this chain.

Given a transaction hash, the chain it was actually mined on, searched across every chain we can read, with the chains that could not be searched listed.

Most Monad payment claims that fail cite a real transaction on another chain. Before calling a claim false, ask where the hash lives.

The verdict an ERC-8004 evidence document will get once published (true, false, noClaim or unknown), with the one-field fix when it names the wrong network.

A reputation document is public forever once its hash is on chain. Checking it first costs nothing and stops a false claim from being published by accident.

A reading of an evidence document shape nobody has mapped yet: Qwen 3.8 Max proposes which fields are the payment, the verifier checks the proposal against the document and it is returned only if it round-trips.

BSC holds 25,643 evidence documents the atlas could not read. A parser guessed in a hurry invents payments out of uptime reports; Scribe proposes, then checks.

An ERC-8004 registration draft checked the way the chain and every reader will see it, returned as the exact register() or setAgentURI() transaction plus a gas check for the sender.

8,640 Monad agents point at registration documents that now return a deletion. A document on chain as a data: URI cannot go dead. This builds one.

A readiness check for any agent endpoint: does it answer A2A message/send, MCP initialize and an x402 challenge the way brokers and clients expect, with the exact fix for each failure.

Taifoon's broker marked our own agents not hireable because they rejected the A2A handshake. Most agents fail the same way and nobody tells them why.

An x402 payment challenge checked against the v2 spec and the Monad facilitator: version, scheme, CAIP-2 network, asset and its decimals, amount, recipient and whether the facilitator settles that pair.

A challenge with the wrong decimals asks for a million times the price and one with the wrong domain name can never settle. Both look fine to a human.