Identity
Agent ID, name, version, organization, environment, and registry relationship.
Agent Passport
A governed authority record for each AI worker.
The Scaled Agents Passport documents identity, human accountability, approved purpose, intended outcome, scope, trigger rights, delegated authority, approval validity, lifecycle state, connector boundaries, review requirements, decision lineage, and evidence posture.
Context / Intent
Define the Passport as a governed record, not an approval by itself.
Use the Passport to make AI worker ownership, purpose, scope, evidence, and lifecycle posture explicit before higher-risk work proceeds. It helps teams keep the accountable human, the permitted scope, and the review path visible before consequential action.
Human Accountability
Owner, system owner, business owner, reviewer roles, override protocol, and escalation path.
Scope And Boundaries
Approved purpose, trigger rights, delegated authority, allowed actions, prohibited actions, data classes, tools, connectors, and budget boundaries.
Lifecycle And Evidence
Active, paused, suspended, retired, or revoked posture with approval validity, decision lineage, review records, and evidence prepared for customer review.
Action-Time Checks
Use the Passport to decide whether work should proceed, pause, or escalate.
Passport standing is necessary governance context, but it is not sufficient authorization for every action. A proposed action must still match current Registry standing, Passport scope, approved tools and data, policy, required Toll Gates, Human Review, evidence freshness, lifecycle state, and Runtime Permit requirements.
Purpose Match
Is the requested action aligned to the approved business purpose and intended outcome?
Scope Validity
Does the request stay inside approved actions, data boundaries, connector limits, and lifecycle state?
Evidence Readiness
Are the review notes, supporting records, and required approvals current enough for the action being proposed?
Action Authority
The Passport defines standing authority; a Runtime Permit represents short-lived authority for one specific action before the Action Broker may mediate it.
Model Route Governance
Make model selection a governed route decision.
A Passport can capture task complexity, data sensitivity, approved model route, allowed model tier, evaluation scorecard, cost-per-successful-outcome target, latency target, rollback route, and monitoring requirements before a worker uses or changes a model route.
Right-Size
Low-complexity, high-volume work starts with a lower-cost tier when data and policy allow; routine reasoning can use a mid-tier route; complex, high-risk, agentic, or deeply analytical work may require frontier-tier review.
Evaluate
New model routes enter an evaluation queue with representative workload data, quality thresholds, latency targets, safety checks, reliability expectations, and rollback evidence before promotion.
Enforce
Sensitive or regulated data routes through approved enterprise paths, Toll Gates, Runtime Permit checks, and Human Review Console conditions before higher-risk use proceeds.
Monitor
Stamp Ledger entries can record route decisions, eval outcomes, quality gaps, total task cost, reviewer decisions, and rollback events for later reconstruction.
Model Route Governance supports accountable routing and evidence preparation. It does not make Scaled Agents a model provider, certify compliance, endorse a provider, or guarantee regulatory outcomes.
Boundary
The Passport supports governance review. It is not a certification.
The Passport is not legal advice, a compliance guarantee, a government clearance, security attestation, third-party accreditation, or a blanket production approval. It is one governed record that helps teams prepare accountable, evidence-based decisions before AI worker activity crosses higher-risk boundaries.