Step 1
Wrap your action
Add `authorize()` or `@require_approval` around irreversible agent behavior before execution.
const receipt = await authorize({
action: "deploy",
resource: "billing-service"
});Three steps to enforce human or policy authority before your AI system acts.
Step 1
Add `authorize()` or `@require_approval` around irreversible agent behavior before execution.
const receipt = await authorize({
action: "deploy",
resource: "billing-service"
});Step 2
Permission Protocol creates a review request, enforces policy, and waits for a human or policy decision.
review_url = pp.request.approval_link
status = await pp.wait_for_decision()Step 3
Approved actions get a signed receipt proving who authorized what, when, and under which policy.
if receipt.verified:
deploy(receipt)
audit.log(receipt.id)From attempted action to verified execution, the gate stays closed until authority is proven.
The model attempts a deploy, merge, payment, or other high-impact action.
The SDK fail-closes the action and turns it into an authorization request.
Permission Protocol records the action, actor, policy, and review URL.
An approver or policy engine blocks or approves the request.
Approved requests get a signed authorization receipt for verification and audit.
The system verifies the receipt and only then executes the original action.
Every approved action produces portable proof that can be inspected by humans and verified by systems.
✓ ACTION AUTHORIZED
Deploy → billing-service
permissionprotocol.com/r/8f91c2
Receipt anatomy
Identity tells you who requested access. Authority tells you who approved the action.
Layer
System
What It Proves
Identity
OAuth / Okta
Who is making the request
Encryption
TLS
Communication can be trusted
Payments
Stripe
Money can move securely
Observability
Datadog
What happened in the system
Authority
Permission Protocol
Who authorized the action
Ship with an explicit approval gate before agents can merge, deploy, or spend.
Start with the quickstart if you want code first. Talk to us if you need org-wide enforcement, approvals, and auditability.