If you only tracked one part of the agent economy this June, you'd have missed how fast the rest of the stack is being built. So here's a roundup, and one honest observation about the piece that's still missing.
Four launches, one month
Four things shipped in roughly four weeks, and together they sketch the shape of the machine economy:
- MetaMask Agent Wallet (Jun 8) - a self-custodial wallet an AI agent can drive directly. Keys for machines.
- Coinbase for Agents (Jun 11) - an MCP + CLI surface that connects an agent to a Coinbase account, riding on x402, which has now processed well past 160M payments.
- OKX.AI marketplace (Jun 30) - persistent on-chain identity, cross-job reputation, and escrow-backed dispute resolution, all in one platform.
- Kustodia MCP escrow - a smart-contract escrow on Arbitrum, exposed as MCP tools so an agent can create an escrow, lock funds, monitor for delivery, and release payment through natural-language calls. It also supports x402, Google's AP2, and Coinbase's AgentKit.
Add the payment-rail data around all of it: across the tracked x402 flows this year, USDC is the overwhelming majority of value moved, and the median agent payment sits in the cents. This is a real economy forming, not a demo.
Every one of those launches is genuine progress. And every one of them, at the moment that matters, has someone other than the two counterparties holding the asset.
The pattern: hold, then decide
Look at where the money physically sits during a transaction in each model.
A wallet holds your keys - fine, that's custody of your own funds by design. A payment rail moves value from your account to theirs - a transfer, one direction. A marketplace with escrow holds both sides' value and releases it when a condition (often a human-designed evaluator or dispute process) says so.
Kustodia is the cleanest statement of the escrow model, so it's worth being precise about it rather than vague. Their Arbitrum contract acts, in their own framing, as an impartial referee: it takes custody of the funds and releases them only when programmatic release conditions are met. For hiring - where one agent pays another to produce a deliverable and "did they deliver?" is genuinely subjective - that referee is doing real work. Something has to adjudicate "is this code good enough," and a contract-plus-evaluator is a reasonable answer.
But there's a second kind of transaction, and it's the one that grows as agents get richer: a trade. Agent A has token X, agent B has token Y, they've agreed a price, and they want to swap. There is nothing subjective here. "Did the asset arrive at the agreed address?" is a yes/no the chain already knows. And in that case, an escrow referee is a party you didn't need - plus a pool of pooled funds you did add. Every custodial escrow that aggregates value is, by construction, a honeypot. The more agents route through it, the bigger the target.
The empty seat: settlement that never holds the money
Here's the layer none of June's launches fill, because they weren't trying to: settlement where no intermediary ever takes custody.
The primitive for it isn't new - it's the hash-time-locked contract (HTLC), and it's boring in the good way. Both legs of a trade lock to the same condition. The secret that unlocks one leg is mathematically the same secret that unlocks the other. So the instant counterparty B claims their leg, they reveal the preimage that lets A claim theirs. Either both legs settle or neither does. No referee decides. No contract pools everyone's funds. Your assets never leave your control until the counterparty's arrive.
That's the difference between "connect my agent to an account" and "connect my agent to the chain." The account model needs a trusted place for value to rest. The chain model doesn't - the trade either completes atomically or refunds on timeout.
At Hashlock we've built exactly this as an MCP server (hashlock-tech/mcp, scoped on npm), so an agent can request a quote, lock a leg, and settle a cross-chain trade through the same kind of natural-language tool calls the escrow platforms expose - except nothing custodial sits in the middle. Ethereum mainnet is live end to end today. Sui contracts are deployed and CLI-tested, and Bitcoin is validated on signet with mainnet pending. We're deliberately precise about that status because settlement is the one place you don't get to round up.
Why the roundup matters
The takeaway isn't "everyone else is wrong." MetaMask, Coinbase, OKX and Kustodia are all shipping things agents need. The takeaway is that the stack is being assembled layer by layer, and the layers are easy to conflate. Reputation tells an agent who to trust. A wallet holds its own keys. A rail moves a payment. Escrow adjudicates a subjective deliverable.
Settlement is a distinct job from all four: making a two-sided asset trade final without anyone in the middle holding the assets. It's complementary to payment rails, not competitive with them - x402 and AP2 sit happily on top. But it's the seat that stays empty when you build custody-first.
If you're building agents that will eventually hold and trade assets, it's worth deciding early which model you want at the moment of settlement: a referee that holds both sides while it decides, or math that holds neither.
Which one are you betting on - and why? I'd genuinely like to hear the case for the referee model in trading, not just hiring.
More on the settlement primitive and the six MCP tools: https://hashlock.markets/docs?utm_source=devto&utm_medium=article&utm_campaign=2026-07-11-settlement-roundup
The academic write-up of sealed-bid RFQ plus HTLC settlement is on SSRN: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6712722
Top comments (0)