Settings
Admin tunables. Writes are admin-gated and audited (sha256-only payloads); reads are scope-scoped.
Active AI model provider
Which LLM the agent uses for grounded answers. Add the matching API key below.
Model for Anthropic
—No models listed for this provider.
Active HRMS backend
Which backend serves HR capabilities. Per FR-26.1 the picker is bounded by the per-backend capability matrix (see HR tab); greytHR is read-only for FBP / reimbursement / expense and the matrix-aware picker hides it where those capabilities are required. Native is dormant.
Autonomy & approval gates
High-stakes actions require approval by default. Toggle one to auto (zero-touch) — switching records a logged risk acknowledgement (FR-31).