The Sherivs platform architecture

Layer detail

Public navigation never links the internal engineering view at /platform. That route stays session-gated.

The first installed runtime on this architecture is Scholar. Removing it would not redefine the platform.

• Identity and governance

  Available today

  People, organizations, and agents are first-class identities. Access is resolved from identity, membership, role, and policy — not from the interface alone.

• Institutional computing

  Available today

  A shared operating environment where an organization's people, records, and workflows live together instead of in disconnected tools.

• Agent orchestration

  In development

  AI workers can be given bounded tasks inside permissions and human oversight. They are not unrestricted administrators and cannot expand their own access.

• Digital-twin runtime

  In development

  Twins are intended to hold identity, state, relationships, and events, then compute on change or schedule. A complete twin runtime is not a finished public product.

• Intelligence and model routing

  In development

  Model access is designed to go through a governed AI gateway so prompts, cost, and safety are not scattered through the product.

• Execution and event processing

  In development

  Work should move as observable events and missions, not as untracked side effects. Event processing is present in parts of the platform and still deepening.

• Data and memory infrastructure

  In development

  Canonical records stay separate from derived intelligence. Predictions must not silently overwrite verified state.

• Observability and assurance

  Available today

  Operators need health, audit, and verification. A public status page reports machine health without leaking internals.