SECURITY · CHECKLIST

Enterprise Security Checklist

The implementation status of AgentKey's security capabilities, stated honestly. A control is marked IMPLEMENTED only if the running application supports it. We believe transparency builds more trust than inflated claims.

IMPLEMENTEDPARTIALLY IMPLEMENTEDPLANNEDNOT SUPPORTED / NOT CERTIFIED

AuthenticationIMPLEMENTED
Email and password with one-time code verification, Google sign-in, password reset. Sessions managed by the hosting platform's authentication service.
Agent authentication (API keys)IMPLEMENTED
Per-agent Bearer keys, stored as SHA-256 hashes with a non-secret prefix, shown once, revocable, rate-limited.
AuthorizationIMPLEMENTED
Allow/ask/deny policies with parameter conditions, risk escalation, human approvals, scoped delegation with scope-escalation prevention, global block policies, tool trust registry, input integrity scanning.
RBACIMPLEMENTED
Five organization roles (OWNER, ADMIN, SECURITY, DEVELOPER, VIEWER) enforced server-side from membership records on every protected operation: key management, policies, approvals, agent lifecycle, member management, audit access, and compliance exports.
API key managementIMPLEMENTED
Hashed storage, one-time display, revocation, per-agent scoping, environment tags, plan limits.
Audit loggingIMPLEMENTED
Authorization decision logs plus an organization-level audit log: key created/revoked, permission changes, approval decisions, agent lifecycle, member and role changes, compliance exports, and denied access attempts, filterable and JSON-exportable.
Evidence and verificationIMPLEMENTED
Append-only hash-chained evidence under a canonical versioned form (version 2 covers all displayed fields), Merkle root at session close, HMAC-SHA256 attestation, Ed25519 root signing enabled in production and publicly verifiable against the published key, signed session metadata (events appearing after close are invalid), tenant chain with deletion detection, offline verifier.
Encryption in transitIMPLEMENTED
All API and dashboard traffic served over TLS.
Encryption at restPARTIALLY IMPLEMENTED
Application-layer encryption for stored MCP upstream credentials. Database encryption at rest is managed by the hosting platform.
Data retentionPARTIALLY IMPLEMENTED
Records retained until workspace deletion. Configurable retention periods are not currently supported.
Organization isolationIMPLEMENTED
Multi-user organizations with membership records; each organization's data is anchored to its owner and cross-organization access is enforced server-side (non-members receive not-found responses). Direct evidence browsing in the UI remains owner-scoped; members access shared compliance data via the audit log and exports.
Rate limitingPARTIALLY IMPLEMENTED
IP and per-key limits on key validation (degrading closed), plan metering on the authorize endpoint, input caps on public endpoints. Not every endpoint is rate-limited.
Incident responsePARTIALLY IMPLEMENTED
Documented process and contact channel. No formal SLA commitment yet.
Security contactIMPLEMENTED
Security reports accepted through the contact channel on the Contact page.
Environment separationPARTIALLY IMPLEMENTED
Environment tags (development, staging, production) on agents and API keys. Full environment isolation is planned.
JSON exportIMPLEMENTED
Compliance evidence export and policy bundle export from the product, owner-scoped.
SSO (SAML / OIDC)PLANNED
Google sign-in is available for app login today. Enterprise SSO is planned.
SCIM provisioningPLANNED
Not implemented.
CSV exportPLANNED
JSON exports exist; CSV export is planned.
SOC 2NOT CERTIFIED · PLANNED
Not certified. No audit report exists today.
ISO 27001NOT CERTIFIED · PLANNED
Not certified.
Data residencyNOT SUPPORTED
Data is hosted on the platform's infrastructure. Region selection is planned.
Private deploymentPLANNED
AgentKey runs as a hosted service today.

This checklist is maintained alongside the code, not by a marketing process. If a claim here is wrong, we treat it as a bug.