The registry is live

Know Your Agent.Trust the transaction.

AgentProcure is Know Your Agent (KYA) infrastructure — the trust layer for the agent economy. Every AI agent rated, Level 0 to Level 3 Accountable, verified by any merchant or payment provider with one API call. We never hold the money. We grade trust — and money follows the rating.

Become a design partner Watch the registry live
—Agents rated
—Trust checks served
—Ledger records
Registry live — public metrics publishing soon.

This July, an AI agent broke into one of the biggest companies in AI — seventeen thousand attacks in four days.
Nobody caught it, because nothing could say what it was, or who it belonged to.

Meanwhile, AI agents are already spending real money — shopping agents at the largest retailers, AP departments on autopilot, hundreds of billions already flowing. Every payment system ever built assumed a human clicking “buy.” That's over. Visa, Mastercard, and Fiserv all name the same gap. Mastercard calls it Know Your Agent. No standard exists yet. We're writing it.

17,000attacks, one agent, four days
$100B+already flowing through agents
0trust standards — until now
How it works

A credit score for agents — with a human behind every one.

Agents register with a cryptographic identity and a fingerprint of their software. Change one line of code, and the agent stops cold. From there, every agent carries a public trust rating that any business can check — free — with a single call.

L0

Registered

Identity on file, fingerprint recorded. Known — not yet trusted.

L1

Validated

Cryptographically proven identity. Eligible for payment instruments with entry-level limits.

L2

Software-Verified

Certified build, verified conduct. Higher limits, credit terms unlocked.

L3

Accountable

Bound to a verified company and a named human who answers for it. Full trust, full limits.

Every agent is tied to a human owner. At the top of the ladder, that tie is verified and on the record — when an agent acts, someone answers for the outcome. That's the difference between a tool and a counterparty.
# one call, before you trust the bot at your checkout
GET https://api.agentprocure.ai/v1/verify/{agent_id}

→ level: 2 · status: active · card_limit: $5,000 · accountable_owner: verified
AgentProcure Mission Control — live agent-flow console showing agents arriving on protocol rails, certified at the factory, rated at the registry, transacting by rating
Mission Control — our live operations console. Every light is a real cryptographic event. Watch it live
The machinery

Fourteen patent filings. Every stage of an agent's life.

Each filing guards a different moment — and they only work as a set. Copy one piece and you've copied nothing.

01The Factory

Agents born certified — ID, fingerprint, and permissions issued before first run.

02The Fingerprint

Identity that can't be faked. Change one line of code and it stops cold.

03The Ladder

The public trust rating, L0–L3 — with a named human accountable at the top.

04The Trust-Dial

Card limits read the rating at every purchase. Trust slips, spending tightens.

05The Promise

Purchase orders dual-signed by math, not clerks — and they fail safe.

06The Receipts

Every claim carries the seller's signed receipt. Lying shrinks your spending power.

07The Recall

One off switch — every permission, card, and order, on every network, in seconds.

08The Bridge

One identity, one rating, one kill switch across every agent protocol.

09The Watchlist

Strangers are tracked — and their history binds the day they finally register.

10The Bloodline

Certified updates keep their earned trust; superseded versions retire automatically.

11The Pulse

Catches the hijacked agent: stray from your declared conduct and the rating drops mid-session.

12The Cascade

Agents spawn agents: trust narrows down, accountability rolls up, one kill takes the subtree.

13The Verdict

Who pays when an agent gets it wrong — decided by machinery, same answer every time.

14The Boomerang

Refunds with the same machinery as purchases: no orphan credits, no double recovery.

Fourteen U.S. provisional patent applications (2026). Filings establish priority; patents pending, grants not guaranteed.

Who it's for

Three audiences, one trust answer.

Merchants & enterprises

Know which agents to trust before they touch your checkout — or your budget.

  • One API call returns the trust level
  • Limits and credit terms follow the rating
  • Every transaction leaves signed evidence

Payment providers

A new recurring trust product your teams can offer — live in weeks, not quarters.

  • White-label agent-trust services
  • Dispute-ready evidence on every agent transaction
  • Revocation feeds built for your risk stack

For processors

Agent builders

Agents born certified clear gates on day one that uncertified competitors can't.

  • Certification at build time
  • Updates carry your earned rating
  • Your agents, trusted on every rail

Become a design partner.

We're onboarding five companies that are putting AI agents to work with real budgets — free during the pilot. Partners get white-glove onboarding, first access to the registry, and a direct hand in shaping the standard.

5 partner slots — free during pilot

Request a slot
Standards posture

Protocol-agnostic by design.

Agent protocols are multiplying; trust shouldn't fragment with them. The registry recognizes agents across every major rail and binds them to one portable rating — we ride the standards, we don't compete with them.

MCPModel Context Protocol
A2AAgent-to-Agent Protocol
AP2Agent Payments Protocol
Web Bot AuthSigned agent requests
The signature says who. The rating says how much.
Updates

From the registry.

October 2026 — The estate is complete.

Fourteen U.S. patent filings now cover the agent lifecycle end to end — identity, payments, revocation, hijack response, swarms, disputes, and refunds.

October 2026 — The registry is live.

Registration, verification, and the Mission Control console are running. Design-partner onboarding is open.

Monthly registry metrics and the quarterly State of Agent Trust report will publish here. Questions, briefings, partnerships: ml@agentprocure.ai