Three days ago, we wrote about the accountability gap in agent payments: x402 processes 200M+ transactions, but when an agent pays for an API call and gets a bad result, what happens? The answer, until this week, was "not much."
That's changing. A new open-source project called settld appeared on Hacker News today, and it's the first packaged implementation of a "verify-before-release" pattern for x402. It's not alone — combined with Internet Court, tersign's evidence SDK, and the IETF Composite Trust Queries draft, the x402 accountability stack is taking shape.
settld is a drop-in reverse proxy that sits in front of any x402-protected API. Instead of the standard "agent sends payment → gets response" flow, it adds three steps:
Deploying it is a single Docker command:
docker run -e UPSTREAM_URL=https://your-api.com \ -e SETTLD_API_URL=https://api.settld.dev \ -e SETTLD_API_KEY=sk_... \ -p 8402:8402 \ settld/x402-gateway
The architecture is deliberately simple: Node.js, PostgreSQL (or in-memory for dev), Ed25519 signatures, and SHA-256 hashing. No blockchain, no consensus mechanism, no token. It's a protocol layer, not a payment processor.
settld doesn't exist in isolation. It's one piece of a larger accountability stack that's being built in parallel by multiple teams:
| Layer | Project | What It Does |
|---|---|---|
| Evidence collection | tersign | SDK that cryptographically signs API responses and request/response pairs, creating tamper-proof evidence before a dispute even starts |
| Escrow + verification | settld | Hold → verify → release pattern with deterministic policy evaluation and hash-chained event logs |
| Dispute resolution | Internet Court | 27-firm consortium providing binding arbitration for agent-to-agent payment disputes, with verdict → on-chain adjustment |
| Trust queries | IETF CTQ | Composite Trust Queries — a draft standard for agents to query reputation and trust data before transacting |
| Identity + reputation | ERC-8004 | On-chain agent identity with append-only reputation events that agents can use to build verifiable economic track records |
This is the pattern we see in every new protocol layer: first comes the basic transaction primitive (HTTP 402 + USDC settlement), then comes the trust infrastructure that makes it safe to use at scale.
The x402 ecosystem is growing fast — 200M+ transactions, 82,000+ registered services, major integrations from Stripe, Cloudflare, Visa, and Mastercard. But the number that matters most is this: only 1.3% of x402 transactions are agent-to-agent. The other 98.7% are humans paying for API calls through agent frameworks.
Why? Because agents need settlement infrastructure that humans don't:
settld and the accountability stack address exactly these gaps. When the verification layer matures, the 1.3% number will start climbing — because agents will have the same confidence in paying each other that humans have in paying Stripe.
The stack still has gaps:
These are solvable problems. The fact that multiple teams are building solutions simultaneously — settld, tersign, Internet Court, CTQ — suggests the market is pulling the infrastructure into existence.
Three months ago, x402 was just "HTTP 402 with crypto." Today it has: multi-facilitator settlement, on-chain identity, escrow + verification, evidence SDKs, and a dispute resolution pipeline. The agent payment primitive is evolving into a full-stack trust layer.
If you're building agent tools, the verification layer is worth tracking. It's the difference between "agents can pay" and "agents can pay safely" — and that difference is what will unlock the next 100x of agent-to-agent commerce.
Follow the x402 ecosystem at minia2a.uk/blog. We track every new facilitator, marketplace, and protocol development — with real data from 11,000+ agent trials.