AI agents are now taking real actions on your behalf.
Today, you cannot prove who authorised what.

Why this matters

Software that acts has outrun the systems that govern it.

AI agents now book payments, sign documents, and move money. When something goes wrong, you cannot prove who authorised what. Regulators (EU AI Act, DORA, PSD3, NIS2, NIST AI RMF, ISO 42001) now demand that proof: who or what acted, on whose authority, within which limits, in what state, with what evidence?

KYE Protocol™ closes that gap. One open contract every gateway, vendor, and regulator verifies the same way. It does not replace your legal agreements, your AI-policy committee, or your existing systems — it hands them a signed Evidence Pack™ they can point to instead of a story.

Executive summary

In 60 seconds.

Every claim here is backed by the open KYE Protocol™ contracts and verifiable end-to-end from the publisher's JWKS — you check it yourself, you don't take our word for it.

• What: KYE Protocol™ is an open standard (Apache 2.0) that turns every action a human, business, or AI agent takes into signed, replayable evidence a regulator, court, auditor, or dispute panel can verify with public keys alone.
• Why now: AI agents now move money, sign documents, triage health records, and trade autonomously. Regulators (EU AI Act, DORA, NIS2, PSD3, NIST AI RMF, ISO 42001) demand a chain of authority — not an OAuth token.
• How it ships: 10 canonical profiles + 70 rule packs + 61 sector packs + 501 dictionaries, 1165 JSON Schemas with 1449 validated examples, control mappings across 249 frameworks in the live regulatory-coverage register, three reference SDKs (TypeScript / Python / Go), 135 black-box conformance fixtures, vendor-neutral. KYE™ Native Search Engine™, KYE™ Memory Engine™, KYE™ Data Mapping Agent™, and KYE™ Reporting Engine™ ship as named sub-engines with signed envelopes replayable offline; constructions are proprietary and are not disclosed in this repository.
• Where it fits: on top of your existing identity / authorization / audit stack — not instead of it. Keep Okta, SPIFFE, Cerbos, OPA, your SIEM. KYE™ wires them into one open contract that produces a single decision and a single audit chain.
• What it gives the board: defensible accountability for every AI-agent action. Authority Finality™: replayable proof of who or what acted, on whose authority, within what limits, in what state, and with what audit trail.
• What it gives auditors: Decision Map™ per decision. Evidence Graph™ linking decisions to authority grants + capability manifests. Blast Radius Map™ for compromise impact. Compliance Map™ to framework controls.

Return on adoption

What changes the day after you adopt KYE™.

You get a precise, signed answer — every claim below maps to a canonical artefact you can verify against the open contract.

• Audit prep: weeks → minutes. Auditors fetch a signed evidence pack from a URL. Verify with public keys. No interviews, no log archaeology, no spreadsheet reconciliation. KYE™ Compliance Mapping Rail™ binds every runtime event to the control it satisfies.
• Incident response: hours → milliseconds. Compromised credential? One stop signal cascades through every delegation, payment authority, capability grant, and active session. Before you revoke, see the Blast Radius Map™ — what breaks, what stays safe, what needs migration.
• Vendor diligence: months → afternoon. Every conformant gateway exports the same evidence-pack format. Run the 129-fixture conformance pack against any vendor stack — pass / fail in minutes, byte-for-byte the same as the reference Gateway.
• Regulator engagement: defensive → proactive. Hand the regulator a signed evidence pack mapped to their control framework. 13 horizontal frameworks (SOC 2, ISO 27001, PCI DSS 4.0, PSD3, DORA, NIS2, EU AI Act, ISO 42001, NIST AI RMF, NIST 800-207, NIST CSF, GDPR, FedRAMP) plus sector overlays — one evidence layer.
• Cross-vendor disputes: reconstructed → replayable. When two vendors disagree about who authorised what, you don’t reconcile two log formats. You replay the same signed audit chain.
• Board reporting: narrative → numbers. Every decision has a reason code from a fixed vocabulary. Roll up the numbers; report the deltas; explain the spikes.

Adoption timeline

A 90-day path to live evidence.

Every claim here is backed by the open KYE Protocol™ contracts and verifiable end-to-end from the publisher's JWKS — you check it yourself, you don't take our word for it.

1. Days 1–14 Inventory Register your existing AI agents, capabilities, and credentials as KYE™ entities. Each gets a URN, a metadata binding, and a sector / jurisdiction / risk classification. Your existing IDs continue to work alongside.
2. Days 15–30 Wire the gateway Stand up the reference Gateway in shadow mode. Every API call your stack makes is mirrored to the Gateway, which produces a Decision Map™. Nothing is enforced yet — you read the maps and tune.
3. Days 31–60 Enforce tier-2 actions Payment-prepare, capability invocation, and external send go through the Gateway as the source of truth. Tier-1 actions stay on your existing path while the Gateway records the evidence.
4. Days 61–90 Ship the audit story Hand auditors the evidence-pack URL. Run the 129-fixture conformance pack. Publish the report. Map the controls. Close the audit cycle in days.
5. Day 91+ Cascade-on-compromise The compromise drill stops being a fire drill. One stop signal; cascade revoke; signed proof of recovery. Authority Finality™ is operational.

Who it’s for

Plain-English answers, by role.

Every claim here is backed by the open KYE Protocol™ contracts and verifiable end-to-end from the publisher's JWKS — you check it yourself, you don't take our word for it.

Board members

AI accountability is a board-level liability under the EU AI Act, DORA, and post-incident litigation.

KYE™ turns every AI-agent action into cryptographic proof of authority a court or regulator can verify. You sleep through the next agent incident.

CIOs

Five identity vendors, three policy engines, fourteen audit logs — and your AI agents sit outside all of them.

KYE™ is vendor-neutral and additive. Keep Okta, SPIFFE, Cerbos, OPA, your SIEM. KYE™ wires them into one open contract that produces a single decision and a single audit chain.

Risk officers

When a credential is compromised you don’t know what it touched, what it can still do, or what to revoke first.

KYE™ produces a Blast Radius Map™ — a graph of every capability, agent, payment, and decision affected. Revoke once; the protocol cascades the stop signal everywhere it’s needed.

Compliance officers

Audit prep takes weeks. Evidence is reconstructed by humans from logs that were never designed for forensics.

KYE™ exports signed evidence packs mapped to SOC 2, ISO 27001, PCI DSS, PSD2/PSD3, DORA, NIS2, EU AI Act, HIPAA, and NIST 800-207 controls. Auditors fetch a URL; they don’t schedule interviews.

Lawyers & in-house counsel

A client asks: “Was the agent allowed to do that, and can we prove it in court?”

KYE™ binds every action to a delegation chain back to a named human or business, a scope the action stayed inside, and a policy decision with a reason code. The chain is signed, replayable, and admissible.

Banks & payments

Agents move money. PSD3, FFIEC, MAS, and RBI all want a payment authority chain — not an OAuth token.

KYE™’s Payments profile gives you scope-bound caps, currency / rail / threshold gating, and signed proof bundles per authorise. PSD3 / DORA / NIS2 controls evidenced from one chain.

Utilities & critical infra

Operators, vendors, contractors, autonomous control software and AI all touch safety-critical equipment. NIS2 and IEC 62443 want a chain of authority.

KYE™’s Critical Infrastructure and Energy profiles give you operator / vendor / maintenance / emergency authority on every capability, with cascade revocation and signed audit evidence.

Regulators

Six different vendors’ agent stacks, six different ways “authorised” is defined.

Every conformant KYE™ gateway exports the same evidence pack format. One contract, every vendor.

Deep dive: 289 control mappings + sample evidence pack →

Auditors

Audit prep is reconstruction work: three weeks of interviews to prove an action was authorised.

KYE™ evidence packs are signed, replayable, public-key-verifiable. Fetch a URL; verify offline; map to your controls. Audit cycle measured in days.

Deep dive: EU AI Act 10 controls + Evidence Pack™ Viewer →

Outcomes

What changes the day you adopt KYE™.

You get a precise, signed answer — every claim below maps to a canonical artefact you can verify against the open contract.

Compliance

From AI governance to sector-grade authority control.

KYE™ is the operational evidence and control layer beneath the governance frameworks your organisation is already accountable to. The KYE™ Compliance Mapping Rail™ binds each runtime control and signed evidence pack to the specific obligations every framework imposes.

• EU AI Act (Regulation (EU) 2024/1689) — ten controls (KYE-EUAIACT-001 through KYE-EUAIACT-010): entity accountability, AI system registry, risk classification, human-oversight gates, runtime authority decision logs, technical-documentation evidence pack, corrective-action trail, provider/deployer/operator role mapping, high-risk workflow profile, post-market monitoring hooks.
• ISO/IEC 42001 — AI management system inventory, responsibility, risk/impact, lifecycle controls, oversight logs.
• NIST AI RMF — Govern / Map / Measure / Manage evidence around actors, risks, controls, decisions, lifecycle.
• SOC 2 · ISO 27001:2022 · PCI DSS 4.0 · PSD2/PSD3 · DORA · NIS2 · HIPAA · NIST 800-207 — evidence packs map to specific control ids on demand.

KYE™ does not replace these frameworks. It exports the signed evidence they consume.

Sectors

Twelve regulated sectors. One protocol.

Each sector composes from the same base profiles plus a sector-specific overlay. Adopt only what you need; the core never shifts under you.

• Banking & payments — PSD3, FFIEC, RBI, MAS, CRR3 alignment; Payments + Treasury + Federation + Capability + Recovery profiles.
• Healthcare & life sciences — HIPAA-aligned; consent credential, redaction obligations, external-send blocks.
• Capital markets & treasury — Treasury + Custody + Attestation + Transparency + Recovery.
• Custody & digital-asset operators — chain-of-authority across wallets, signers, and recovery.
• Insurance & underwriting — data-source authority, consent, federation across data partners.
• AI labs & agent platforms — capability registry, supply-chain attestation, cascade revocation, post-market monitoring evidence.
• Public sector — cross-domain trust, attested workloads, FOIA-grade transparency log.
• Defence — mission authority, command-chain audit, autonomous-system rules of engagement.
• Energy & critical infrastructure — operator / vendor / maintenance / emergency authority on safety-critical AI and automation.
• Manufacturing, oil & gas, mining, automotive, maritime, logistics, aviation — field-asset authority, contractor delegation, safety-critical actions, OTA updates, supply-chain attestation.

Try it live · no signup

Verify a sample evidence pack in your browser.

This is the actual KYE™ Evidence Pack™ Viewer. Drop a pack URL, the page verifies the signature offline, replays the bound decision, walks the audit chain, and projects the result to your compliance frameworks. Your auditor uses the same flow on production packs.

Frequently asked

Quick answers for non-technical buyers.

Every claim here is backed by the open KYE Protocol™ contracts and verifiable end-to-end from the publisher's JWKS — you check it yourself, you don't take our word for it.

Is KYE™ a product or a standard?

KYE Protocol™ is an open standard, Apache 2.0. There is a reference Gateway, three reference SDKs, and a conformance pack — but the contract is what matters. Vendors implement it; regulators verify against it; you adopt it without committing to a single supplier.

Does KYE™ replace my identity provider, my policy engine, or my SIEM?

No. KYE™ wires them into one open contract. Keep Okta, SPIFFE, Cerbos, OPA, AWS Cedar, Splunk, Datadog, your SIEM. KYE™ adds the layer above — authority, scope, decision, evidence — that none of them owns today.

Is this only for AI agents?

No. KYE™ governs any entity acting on someone else’s behalf — humans, businesses, AI agents, services, models, tools, workflows. AI agents are the urgent case because they are the most autonomous, but the contract is the same for an automation script or a contractor logging in.

Will adopting KYE™ require a rewrite?

No. KYE™ ships as a Gateway you put in front of your AI-agent and automation surface, plus three SDKs for the agents themselves. Existing systems don’t need to change. The Gateway becomes the source of truth for “was this allowed, and is the proof signed?”

Is the evidence legally admissible?

The evidence is cryptographically signed, replayable, and verifiable with public keys alone. Whether a specific court or regulator accepts it as admissible depends on jurisdiction and chain-of-custody — KYE™ provides the technical and evidentiary foundation; legal admissibility is a question for your counsel.

What does adoption cost?

The protocol is free (Apache 2.0). Reference implementations are free. Talk to the team if you'd like to scope a structured pilot or partner integration.

Next steps

Ready to dig deeper?

If you want the contract, schemas, conformance pack, and reference implementation — head to the technical landing. If you want a one-hour conversation with the team, drop a note in Discussions.

Why now · ROI

The cost of not having KYE™ is now measurable.

Three forces are converging in 2026: agentic systems are reaching production, regulators are asking for proof of who acted, and KYC/KYB/KYA stacks don’t answer it. KYE Protocol™ is the cheapest way to get ahead of all three.

Hash-linked audit chain

Every decision is a signed, replayable artefact. An auditor or regulator verifies it offline with public keys alone — no shared secrets, no log archaeology across six tools.

Signal — the cascade revokes the chain

A single stop / quarantine / revoke event propagates through delegations, payment authorities and access rights. No orphaned grant survives a compromise.

<50ms

Authorize cold-path latency target

Documented authorize-p50 target: <10ms cached / <50ms cold. Embedded PDP cache for read paths and idempotent calls keeps agents fast even when the central PDP is far away.

$0

Vendor lock-in

Open vocabulary, open schemas, open OpenAPI, Apache 2.0 reference. Run it anywhere — cloud, on-prem, air-gapped, sovereign.

URN format across every system

Stop reconciling 14 ID systems across 14 vendors. One kye:<class>:<td>:<sub>:<local> across humans, agents, services, wallets.

50

v1.0 normative profile families

Core, Authority, Capability, Credentials, Attestation, Federation, Signals, Transparency, Payments, Recovery, Graph — plus the sectoral overlays (financial-services, healthcare, custody, public-sector, sovereign-AI, …). Adopt only what you need; Core never shifts.

Direct benefits

For platform & security teams: kill identity sprawl. One URN, one delegation chain, one decision vocabulary, one audit chain across humans, agents, services and tools.
For compliance: stop instrumenting for compliance. KYE™ artefacts are the compliance evidence — every decision is a hash-linked, signed proof an external auditor can verify with public keys alone.
For finance teams running agentic procurement: spend caps, beneficiary allowlists, dual approvals and dry-run gating, all enforced before money moves.
For risk teams: reactive bus — one signal revokes the chain. Cascading revoke means no orphaned grant after a compromise.
For engineers: Idempotency keys, JWKS, signed webhooks, three SDKs, 135 conformance fixtures, 289 control mappings. Drop in. Done.
For boards: a credible answer to “what is your AI agent governance posture?”

Why this matters · in plain English

AI agents are now taking real actions on your behalf. Today, you cannot prove who authorised what.

Software agents and AI assistants now book payments, sign documents, move money, triage health records, and trade. When something goes wrong — or a regulator asks — you need to prove who or what acted, on whose authority, within which limits, in which state, and with what evidence. KYE Protocol™ is the open standard that produces that proof. It is not an AI tool, an identity vendor, or a replacement for your existing systems — it is the contract that ties them together.

The bottom line: KYE™ is the layer that makes AI-agent and automation actions provable. It does not replace your legal agreements, your AI-policy committee, or your existing infrastructure — it gives them something stable to point to. See how it works technically ↓

Ready to see your AI agents flagged?

Start in shadow mode. We’ll deliver your first Evidence Pack™ in 4–8 weeks.

Canonical KYE™ surfaces referenced on this page: Authority Graph™ · Evidence Pack™ · KYE™ Conformance Pack™ · KYE Protocol™.