Licensed Platform
Scaled Agents provides the product, governance patterns, and implementation materials. The customer determines the approved hosting path, operating model, owners, and review gates.
Deployment And Data Control
Customer-Managed License
Scaled Agents is designed for deployment inside customer-managed environments, including private cloud, on-premises, and controlled preview environments. Registry data, Passport records, telemetry, and evidence should remain customer-controlled unless a separately reviewed operating model says otherwise.
Plain English
Customer-managed means the control plane runs within the customer's approved boundary.
The license model is intended for organizations that want AI worker governance records, evidence, operating controls, and connector decisions to stay under their own environment, review process, and accountable ownership model.
Customer-Controlled Records
Passport records, evidence, registry posture, telemetry summaries, and connector inventories should remain inside customer-approved control boundaries unless a separately reviewed model says otherwise.
Support Boundary
Advisory support can help teams plan, configure, and review the posture. It does not replace customer security, privacy, legal, compliance, architecture, or production approval decisions.
Context / Intent
Explain the intended deployment and data-control posture.
Use this page to orient customer, security, architecture, and governance teams before any implementation decision. It frames the intended control boundary so teams can ask the right ownership, environment, and evidence questions early.
Customer-Managed By Default
The default model is licensed enterprise software installed and operated by the customer, with customer control over registry data, Passport records, evidence, retention, access, monitoring, and exception workflow.
Preview Hosting Boundary
Cloudflare Pages or other preview hosting may support review and planning. Preview hosting does not convert the default production posture into a vendor-operated service model.
Evidence Control
Evidence records should stay inside customer-approved control boundaries and avoid secrets, raw payment identifiers, unnecessary personal data, or confidential workflow payloads.
Review Before Expansion
Higher-risk deployment modes, live writes, destructive actions, or deeper connector authority require tested implementation and a separately approved authority model.
Integration Controls
Start with bounded integrations and expand only through approved review paths.
Customer-controlled deployment works best when integration decisions stay explicit. Begin with read-only posture, document connector boundaries, and escalate higher-risk actions to named owners before those actions move into operational use.
Read-Only First
MCP, API, and connector inventory should begin read-only until security and owner approval is recorded.
No Destructive Tools By Default
Write, lifecycle, payment, or production-adjacent actions require tested implementation and approved authority.
Data Minimization
Logs, evidence, and support artifacts should prefer customer-safe metadata, redacted summaries, and non-secret identifiers.
Human Review
High-risk exceptions should route to customer-designated owners before the organization proceeds.
Boundary
Deployment planning describes posture. It does not authorize production operation.
This page helps teams understand the intended architecture and review sequence. It does not provide legal advice, compliance approval, security authorization, production authorization, or proof that a customer environment already satisfies required controls.