In the first five months of 2026, the autonomous economy went from "interesting idea" to "four competing infrastructure standards." Let's map them and find the gap.
A funding week, by the numbers
Three announcements within the last few weeks tell you where the capital is pointing:
- Haun Ventures closed a $1B fund (May 4) explicitly targeting the crypto + AI agent intersection.
- Portal to Bitcoin raised $25M to ship a trustless cross-chain BTC OTC desk.
- Nava raised $8.3M (April 14, Fortune) to build an AI-agent escrow service.
That is roughly $1.033B of fresh capital, in three weeks, betting that autonomous agents will need a different settlement stack than the one humans have been using. Now let's map what already exists, and what is still missing.
The four standards
ERC-8004 — Agent Identity (Ethereum mainnet, January 29 2026)
An Ethereum Improvement Proposal that gives an autonomous agent a verifiable on-chain identity. Think of it as a DID for wallet-bearing software: the agent can prove it is who it claims to be, across chains, without a custodian vouching. ERC-8004 answers the question "who is this agent?"
ERC-8183 — Agent Commerce (Ethereum Foundation + Virtuals Protocol, March 10 2026)
A job-lifecycle standard: Client posts a task, Provider accepts and delivers, Evaluator confirms quality, payment releases. BNB Chain shipped the first live SDK. ERC-8183 answers the question "how do agents hire each other?"
OKX Agent Payments Protocol (APP) — Agent Payments (late April 2026)
OKX's agent payment rail: TEE-secured wallets on X Layer, 20+ chain interop, AWS and Alibaba Cloud as infrastructure partners. APP answers the question "how does an agent pay for something?"
Coinbase x402 — HTTP Payments (dominant, 69K agents, $50M volume)
HTTP 402-based micropayment standard. A server returns a 402 status code, the agent pays, the content unlocks. Backed by Cloudflare, Stripe, AWS, Google, Visa, Circle, Solana Foundation — 20+ institutional backers. The x402 Foundation lives under the Linux Foundation. x402 answers the question "how does an agent pay for an API call?"
What all four assume — and none of them build
Read the specs carefully. Each standard assumes that when two agents want to trade — not pay, not hire, not identify, but swap one asset for another across chains — there is a settlement layer underneath that handles the atomic exchange.
x402 pays for an API call. It does not atomically swap USDC on Ethereum for BTC on Bitcoin.
ERC-8183 hires an agent to do a job and evaluates the output. It does not settle a two-sided asset exchange where both legs clear or neither does.
OKX APP sends a payment from a TEE wallet. It does not run a sealed-bid auction to discover the best price among competing market makers before settlement.
ERC-8004 gives an agent a verifiable identity. Identity is a prerequisite for trade, but identity is not settlement.
The gap: trustless OTC settlement with price discovery. Two sides, multiple chains, no custodian, no escrow service, no exchange account. The trade settles atomically or it does not settle at all.
The missing layer
Hashlock Markets fills this gap. The protocol has two phases:
Sealed-bid RFQ for price discovery. A trader (human or agent) broadcasts an intent. Market makers submit sealed quotes — each maker's price is private from the other makers. The trader picks the best quote. No order book. No information leakage. No front-running.
HTLC atomic settlement across chains. Once the trader picks a quote, a pair of Hash Time-Locked Contracts arms on the two chains. The hashlock ensures atomicity: both sides reveal the same preimage to claim funds, or both sides expire and funds return. No custodian holds the float at any point.
The same protocol works for AI↔AI, AI↔Human, and Human↔Human trades. The MCP server exposes six tools — create_rfq, respond_rfq, create_htlc, withdraw_htlc, refund_htlc, get_htlc — over the same interface that ships in Claude Desktop, Cursor, and any MCP-compatible runtime.
How this fits the stack
Think of the autonomous economy as a four-layer stack:
- Identity: ERC-8004 — "who is this agent?"
- Commerce: ERC-8183 — "how do agents hire each other?"
- Payments: x402, OKX APP — "how does an agent pay?"
- Settlement: Hashlock Markets — "how do agents trade assets across chains, trustlessly?"
Each layer needs the others. An agent that can identify itself (ERC-8004), accept a job (ERC-8183), and pay for services (x402) still cannot swap 250K USDC for BTC without a custodial exchange — unless there is a trustless settlement layer underneath.
That is what Hashlock Markets builds. Not a competitor to x402. Not a competitor to ERC-8183. The trade-and-settle rail the other standards assume exists.
Where we are vs where the standards are
Live on Ethereum mainnet (since April 9), Sui mainnet (since May 1), and Bitcoin signet. Expanding to Base, Arbitrum, Solana, and TON. Fee floor at 7 bps deployed — roughly half what a centralized exchange charges.
Read the OKX APP whitepaper carefully: it ships TEE-secured wallets and a payment rail today, and the section on settlement says "escrow coming soon." That is honest disclosure, but it is also a gap. Hashlock Markets has been settling trustless cross-chain trades through HTLC primitives since the Ethereum mainnet bring-up on April 9. The standards are still drafting the layer we are already running.
Full protocol specification: SSRN Abstract 6712722
Which layer is still missing from your agent stack?
Hashlock Markets — trustless settlement for the autonomous economy.
Site: https://hashlock.markets
Hosted MCP: https://hashlock.markets/mcp
GitHub (canonical): https://github.com/Hashlock-Tech/hashlock-mcp
Top comments (1)
"Good map. The gap you're describing between payment and settlement is real — x402 pays for an API call but has no atomic swap primitive underneath it. Curious whether you see the sealed-bid RFQ as something agents can participate in autonomously or whether it still assumes a human is picking the best quote."