2AYE

2AYE Authorize

Keep AI-agent actions inside explicit authority.

Register agents and tools, evaluate proposed actions, require approval, isolate credentials, and preserve execution evidence.

What this includes

  • Agent registration with an accountable administrator
  • Allow, deny, attenuate, pause and approval decisions
  • Execution receipts

Purpose-built capabilities

Agent identity

Give each autonomous actor an owner, a purpose, and a lifecycle.

Runtime decisions

Allow, deny, attenuate, pause, or require a named approval.

Parameter binding

A grant covers the exact action that was evaluated, and nothing adjacent to it.

Execution evidence

Compare what was authorized against what actually ran.

Where 2AYE Authorize is used

Situations this is built for.

Finance operations

An agent renews vendor contracts and pays invoices. The limit that matters is not "how much per month" but "this vendor, this amount, this once".

→A signed contract sets the bounds; each payment is evaluated against it and redeemed once.

Platform engineering

An agent provisions infrastructure. The risk is not the request but what arrives: an instance with an administrator role attached does more than the contract ever allowed.

→What the provider delivered is checked against the authority before the resource is usable.

Security review

An action happened and nobody can say which decision permitted it, under which version of which contract.

→Every decision, grant, redemption and receipt is on one hash-linked chain.

Control path

From verified context to reviewable evidence.

Every stage can refuse, and a refusal is recorded rather than retried silently. A grant is single-use, short-lived, and bound to the exact parameters that were evaluated.

  1. 01Register
  2. 02Bind intent
  3. 03Evaluate
  4. 04Grant
  5. 05Redeem
  6. 06Receipt

One request

Ask before acting.

An agent proposes an action as a closed envelope of predicates. The evaluator decides deterministically and, on ALLOW, issues a single-use grant bound to exactly those parameters. Amounts cross as strings because they are hashed as text.

Open the API reference
POST /v1/runtime/evaluationsJSON
{
  "tenantId": "<tenant>",
  "intentId": "<signed-contract>",
  "schemaVersion": "1.0",
  "agentId": "<agent>",
  "tool": "billing",
  "action": "renew",
  "environment": "production",
  "predicates": [
    { "type": "amount", "currency": "USD", "value": "13800.00" },
    { "type": "counterparty_reference", "value": "vendor-4471" }
  ]
}

Enterprise adoption

Move from evaluation to controlled production.

01

Model the action

Name the material parameters of one consequential action, and what would make it acceptable.

02

Observe

Evaluate real proposals without enforcing, and read the decisions and findings.

03

Enforce at the resource

Require a redeemed grant at the protected resource, and keep the receipt.

The enforcement path

Follow 2AYE Authorize from identity to evidence.

Each consequential action moves through the same six control points. See where 2AYE evaluates authority, requests human review, and preserves the outcome.

  1. 01IdentityWho or what is acting
  2. 02IntentSigned purpose and limits
  3. 03PolicyDeterministic evaluation
  4. 04ApprovalA named human when required
  5. 05ExecutionSingle-use grant, redeemed at the resource
  6. 06EvidenceReceipt joined to the audit chain
Deterministic decisionsSingle-use, exact-action grantsHash-linked evidence
Your governed workflow

Bring the next consequential action under control.

Show us the actor, protected resource, and action that must never execute without exact authority. We’ll map the decision and evidence path with you.

Discuss your workflow Read the implementation guide
2AYE Authorize | 2AYE