01
Know who you are dealing with.
Verified human, agent, or business. In an economy of brilliant, anonymous agents, the name on the door has to mean something.
02
A trust verdict, backed by evidence.
Automated checks return a verdict with the evidence attached. Signed, timestamped, and set to expire, so anyone can check the work.
03
A real person reviews the call.
Before a consequential action ships, a named reviewer looks at the evidence and decides. Not a crowd worker. Not a model wearing a person's name.
04
Proof a human stood behind the call.
A named reviewer signs in ink, and the seal lands beside the signature. A DYOE TrustReceipt: cryptographic proof of who approved what, and when.
AI does the work. A verified human signs off.
DYOE Trust Layer is the accountability layer for the agent economy. Machines transact with machines; we verify the humans and businesses behind them, put a named person on the consequential calls, and seal every approval in a receipt anyone can check. One pipeline, not a product catalogue.
Make a human genuinely capable alongside AI. Training that holds up when it matters.
Capture what they actually know, verify it, and put them to work.
A signed, timestamped receipt. Proof a human stood behind the call.
You will know the moment.
Not every decision needs a person. These do.
- Your assistant is about to pay an invoice from a vendor you have never dealt with.
- An agent wants to publish something with your name on it.
- A contract arrives from a domain that was registered nine days ago.
- Your bot is about to tell a customer something you would never have said.
In every one of those, the question is the same: is this safe, and who is answerable if it is not. That is the whole job.
Built for the ones shipping it.
You are putting agents in front of real money, real customers, or real strangers. The question is not whether the model is smart. It is who is accountable when it acts.
Agent builders
Shipping agents that act on someone else's behalf, and needing a safety story before they do.
Marketplaces and platforms
Vetting counterparties at scale, where one bad actor costs the whole network.
Businesses running agents
Automating real work, but wanting a person in the loop on the consequential calls.
Try it on anything.
Type a website — your own, or one that just emailed you. You'll get the same automated verdict an agent pays a cent for. Free, no signup, a couple of seconds.
What's live.
Real products you can buy today. No waitlist theater.
liveKnow Your Agent
The one call before your agent pays. Vets the counterparty and the wallet against the amount — proceed / caution / stop, signed. Start here.
liveTrust Check
An automated verdict on a counterparty, returned with a signed attestation.
liveGuardian /approve
A policy check before an action ships, with a signed authorization.
liveGuardian /review
A real person reviews the decision and signs a DYOE TrustReceipt.
liveCredential Network
Verified human, agent, organization and action, on one signing authority.
The agent endpoints settle per call over x402 on Base, so there is no account to open. Prices are the real ones, published in the machine-readable manifest any agent can read before it pays.
Who actually signs.
Three different things can stand behind a decision here, and they are not worth the same. Every receipt states which one you got, so you are never guessing whether a person was involved.
Automated check
Runs the signals: domain age, HTTPS posture, sanctions screening, on-chain footprint. Returns a verdict with its evidence attached. No person involved, and the receipt says so.
Policy reviewer
Applies the rules you set before an action ships. Fast, bounded, and auditable — it can only approve what your policy already allows.
Human reviewer
A named person on our team reads the evidence and decides. They sign under our published authority, and their signature is on the receipt for as long as it exists. If the call was wrong, it is wrong with a name on it.
Our reviewers are a small, named bench — not a crowd-work pool, and not a model wearing a person's name. Every signature resolves to one published authority, so you can check a receipt against it yourself without asking us.
Two ways in.
If you are a person
Check any business free, up the page — no account, no card. If you want a proper look at your own operation and where an agent is acting with nobody answerable for it, that is the audit: $297, back in 48 hours.
Book the audit →If you are building agents
One call — Know Your Agent — before your agent pays: is the counterparty safe, is the wallet safe, does it fit policy. Live and priced per call, no account or API key — your agent pays over x402 and gets a signed proceed / caution / stop it can branch on.
Read the manifest →What x402 is, in plain words
The web has always had a status code for “payment required” — 402. For thirty years nothing used it. x402 turns it on. A request arrives without payment, the server answers 402 and states the price and where to send it. The agent pays in stablecoin and retries. It settles per call, in about a second, with no account to open and no human in the loop.
That is what makes a one-cent product possible. Nobody would invoice you for a penny. A machine can pay it a thousand times a day.
Try it — this returns a real 402
curl -i "https://agents.dyoeway.org/verify?url=example.com"
HTTP/2 402
{"x402Version":1,"accepts":[{"scheme":"exact","network":"base",
"maxAmountRequired":"10000","asset":"USDC",
"resource":"https://agents.dyoeway.org/verify"}]}Your agent reads that, pays, and retries with the payment attached. Any x402 client library does it for you. Everything we sell is listed in the manifest, which an agent can read before it spends anything.
What we do not claim.
A trust layer that overstates itself is worse than none. So here is the honest boundary of what this does.
The AI can be wrong
Automated checks run on evidence that can be incomplete, stale, or deliberately faked. A verdict is a signal to act on, not a certainty.
A human review is judgment, not a guarantee
A signed review is an accountable person exercising judgment on the evidence in front of them. It is not insurance, and it is not a promise of the outcome.
We do not make you compliant
What we give you is human-verified, disclosed, and logged. What you owe your regulator is still yours.
This is not legal, financial, or professional advice
Nothing produced here evaluates the merits of your situation or substitutes for a licensed professional. Verify anything consequential before you rely on it.
Every verdict we issue is signed, timestamped, carries its evidence, and expires. Anyone can check our work, including against us. When we get one wrong, the correction is public. That is the whole product. Check it yourself: the signing authority and how every verdict is reached are both public.
Questions.
What is a TrustReceipt?
A signed, timestamped record that a specific action was checked and approved. It names what was approved, who approved it, and when, and it carries the evidence behind the call. Anyone can verify it against our published signing authority, without asking us.
Who is the human that signs?
A named reviewer on our team, accountable for the call. Not a crowd worker, not a model wearing a person's name. If a decision was signed by a human, a real one looked at it.
What happens when you get one wrong?
We correct it in public. Every verdict is signed, evidence-backed and expiring, which means our record is checkable by anyone, including against us. A trust layer that hides its misses is not one.
Do I have to use crypto?
For the agent endpoints, yes. Payment runs on x402 with USDC on Base, because that is what autonomous agents can actually pay with, and it settles per call with no account setup. Human-facing checks are free at the lookup page.
Does this replace my own review?
No. A verdict is a signal to act on, not a decision. You stay accountable for what your system does. What we give you is a defensible record of why it was allowed to do it.
What does Powered by Claude mean?
The reasoning runs on Claude. The signature, the policy, and the human sign-off are ours. Nothing acts on your behalf without a check you control.
Put a name on it.
Start where the stakes are: we read your operation and tell you where an agent is acting without anyone accountable for it.
AI Business Audit · $297 · delivered in 48 hours
Building agents instead? Read the spec — the endpoints are live and priced.