Every Sunday we take the week's agent-economy news and try to do one useful thing with it: not rank it, not hype it, but place it on a map. Most of these announcements are real and most of them matter — but they do not all live at the same layer, and treating "agent commerce" as one undifferentiated bucket is how builders pick the wrong primitive.

Here is the week, in the order of how much it moved the map.

1. Atomic OTC got funded

The biggest signal this week was capital, not code. A team raised $25M and launched an atomic OTC desk for large cross-chain trades, built on HTLCs and Bitcoin Taproot, with no custodian holding the assets mid-trade.

We want to be precise about why this matters to us, because it is easy to read it the wrong way. It is not a threat to be spun. It is category validation. For the better part of a year, "trustless cross-chain settlement" was a phrase you mostly heard from people building it, including us. When a fund underwrites a desk whose entire pitch is instant, trustless, large-scale cross-chain settlement, the thesis stops being a thesis. The market has now priced it.

Where the map gets interesting is the shape of the bet. A BTC-centric atomic OTC desk optimizes for one thing extremely well: large bilateral trades where Bitcoin is one leg. That is a great wedge. The differences worth understanding, layer to layer:

• Sealed-bid vs open quotes. An OTC desk that shows a quote leaks information — counterparties learn your size and direction. A sealed-bid RFQ keeps the order private until it clears, which matters more, not less, when the trader is an autonomous agent that can be probed at machine speed.
• BTC-centric vs multi-chain native. A Taproot-anchored desk is excellent for Bitcoin legs. An agent that needs to settle ETH↔SUI↔BTC interchangeably needs the same primitive to span all of them.
• Desk vs primitive. A desk is a venue you route through. A settlement primitive is something an agent calls directly, with no desk in the path.

None of these make the funded desk wrong. They describe a different position on the same map — and a funded neighbor is the best evidence yet that the map is worth drawing.

2. x402 crossed 75M transactions

x402 reported 75 million transactions in the last 30 days, and Agent.market — an app store where agents discover and pay each other in stablecoins — is now live on top of it. EVM and Solana. This is the clearest "default rail" trajectory in the space.

The relationship to atomic settlement is the one we keep drawing because it keeps being misread: x402 answers "how does my agent pay." It is an HTTP-native payment handshake, and it is very good at it. It does not answer "how do two strangers swap native assets atomically, across chains, without one of them trusting the other or a custodian in the middle." Those are different sentences. A payment rail moves a known asset to a known recipient. A settlement layer clears a trade between two parties who don't trust each other. You can — and probably will — want both.

3. AP2 went to the FIDO Alliance

Google's Agent Payments Protocol (AP2) was donated to the FIDO Alliance, with 60+ partners now in the orbit (PayPal, Mastercard, Amex, Coinbase, Salesforce). AP2 is about mandates — how an agent proves it is authorized to transact on a human's behalf. It is the consent-and-intent layer.

On the map, AP2 sits above settlement. It says who is allowed to do the trade. It says nothing about how the trade clears without counterparty risk. An AP2-authorized mandate still has to settle against something. That something can be a custodial rail, or it can be a trust-minimized atomic swap. Complementary, not competitive — and the fact that a standards body is now stewarding the consent layer is good for everyone building underneath it.

4. The ERC-8004 / ERC-8183 / x402 stack keeps forming

The Ethereum Foundation and Virtuals Protocol continued pushing ERC-8183 (trustless agent transactions), which slots next to ERC-8004 (agent identity/reputation) and x402 (payments) into what is starting to look like a canonical agent-commerce stack: who the agent is, how it pays, how it transacts with confidence.

There is a hole at the bottom of that stack, and it is the one we care about: when the transaction is a cross-chain swap of native assets between two agents, what actually clears it without a custodian or a bridge honeypot? That is the settlement primitive. ERC-8183 standardizing the "transact with confidence" envelope is a tailwind for us, not a competitor — it gives our settlement layer a standard interface to plug into.

5. Wallet-side MCPs kept shipping

Base MCP — letting ChatGPT, Claude, and Cursor drive Base wallets directly — is still trending. It is genuinely useful: if your agent already trusts a venue, a wallet-side MCP is the shortest path to action. It is exchange-leaning and custodial-adjacent by design. Our MCP server exposes a different primitive: not "operate my wallet on this venue" but "settle a trade with a stranger, trustlessly, across chains."

What we shipped this week

Building in public means the recap includes our own diffs, not just everyone else's:

• Compute-capacity MCP tools. We shipped tooling that lets an agent trade compute capacity the same way it trades tokens — settling a block of compute atomically against payment. An agent that can't pay for the compute it runs on is a stranded asset; this closes that gap. It is the sixth settlement primitive we've built on the same HTLC + sealed-bid RFQ core.
• Surface work. Our /methodology and /terms pages are now in the sitemap, so the technical explanation of how settlement and volume accounting actually work is properly discoverable.

Chain status, stated honestly, every week: Ethereum mainnet is live end-to-end. Bitcoin HTLCs are signet-validated, mainnet pending. Sui contracts are deployed and CLI-tested, with gateway wiring in progress. We don't call Sui or BTC "live" until they are. Rails ready, more trains coming.

The map, in one paragraph

Identity (ERC-8004) says who. Consent (AP2) says allowed. Payment rails (x402) say paid. And underneath all of it, when two agents that don't trust each other need to swap native assets across chains, a settlement primitive says cleared — atomically, with no custodian and no bridge. That last layer is where Hashlock lives: sealed-bid RFQ fused with HTLC atomic settlement. Your money never leaves your wallet until theirs arrives.

A funded atomic OTC desk this week was the strongest evidence yet that the bottom of this stack is real and worth capitalizing. The open question for builders is which layer you're standing on.

Coming Monday: a deep dive on multi-leg trade atomicity — what changes when a single agent trade has three legs that all have to clear or none do.

Where are you building — the payment rail, or the settlement primitive underneath it? I'd genuinely like to know what you're running into.

Read how the settlement and volume math works: https://hashlock.markets/methodology?utm_source=devto&utm_medium=article&utm_campaign=2026-05-31-weekly-meta-recap

The MCP server and tools: https://hashlock.markets/docs?utm_source=devto&utm_medium=article&utm_campaign=2026-05-31-weekly-meta-recap

If you want the academic version of the argument, the whitepaper is on SSRN: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6712722