agents.keyra.ie
Security & trust

Eleven controls. Every deployment.

Keyra agents operate inside regulated environments. The trust model below is enforced on every customer tenant, every subscription, and every agent instance.

01
Tenant isolation

Every deployment runs inside a customer-specific tenant world. No agent ever crosses tenant boundaries.

02
Agent ID traceability

From Ciright parent → Keyra catalog → subscription → tenant instance, every agent is fully traceable.

03
Role-based access

All operations are gated by roles. Operators, reviewers, sovereign admins, and tenant admins have distinct capability scopes.

04
Human approval

Material actions require human-in-the-loop approval. Approval is recorded for every decision.

05
Audit logs

Append-only audit log on every action. Sovereign deployments include ministry-level audit packaging.

06
Permission boundaries

Permission templates are bound at deployment time. Agents cannot self-elevate.

07
API access controls

Outbound API calls are scoped to required integrations. No external network access by default.

08
Knowledge pack controls

Knowledge packs are explicit, versioned, and revocable. Customer-private knowledge stays in the tenant world.

09
Legal approval

Each subscription carries a legal approval status that must be set to Approved before activation.

10
Security approval

Each subscription carries a security approval status that must be set to Approved before activation.

11
Revocation

Subscriptions are fully revocable. Revocation cascades to all tenant agent instance IDs.

Sovereign deployments

For ministry-level and country-resident deployments, Keyra operates under a dedicated sovereign control envelope: in-country infrastructure, sovereign-admin role, regulator-visible audit, and contract-grade billing.