Built to pass the security review.
A coworker that takes real action needs to earn real trust. Every layer is built for the people who have to sign off on it — on-premise deployment, document-level RBAC, risk-tier governance, and an audit trail for every decision.
Defense in Depth
Security built in, not bolted on.
Every component is designed for an agent that reads sensitive data and takes consequential action.
Governance & policy engine
Every agent action carries a risk tier from T0 to T5. Policies resolve at platform, team, and agent levels, and consequential writes pause for human approval before they happen.
- 6 risk tiers (T0 unrestricted → T5 blocked)
- Cross-surface approvals (Slack, email, web)
- Egress controls for external services
- Control Tower admin dashboard
On-premise deployment
Run the coworker on your own infrastructure — AWS, GCP, Azure, on-prem, or fully air-gapped. Your data never leaves your network.
- Docker & Kubernetes ready
- Air-gapped deployment support
- No outbound data transfer
- Bring your own LLM — including local models
Fine-grained RBAC
Four-tier team roles plus document-level permissions, synced from your identity provider — so the coworker only ever sees what each person is allowed to.
- Owner, Admin, Editor, Viewer roles
- Document-level access control
- Permission-aware retrieval
- Synced from Slack, Microsoft, Google
Audit logging
Every search, every access, and every agent action is logged and queryable — with the full reasoning trace behind each decision.
- Complete access history
- Per-run reasoning trace logging
- Tool execution with redacted params
- Configurable retention policies
Input & output guardrails
Multi-layer protection: injection detection, PII filtering, hallucination detection, and source verification. Tool-result outputs are guardrail-checked before the agent can incorporate them — so a database row or API response containing PII is redacted before it reaches the model.
- Prompt injection detection
- PII detection & redaction on input and output
- Hallucination detection (lexical + semantic)
- Tool-result output guardrails before ingestion
- Per-team configurable rules
Identity federation
Map users across Slack workspaces, Azure AD, Google Workspace, and OIDC providers — for permission-aware behavior on every surface.
- Slack workspace → user mapping
- Azure AD identity federation
- Google Workspace integration
- OIDC SSO (Keycloak, Auth0, Okta, Azure AD, Google)
- SAML available on Enterprise plans
Compliance
Meeting enterprise requirements.
SOC 2 Type II
Working toward SOC 2 Type II certification for our cloud offering.
GDPR Ready
Data processing agreements, right to erasure, and data portability support.
HIPAA Eligible
On-premise deployments can meet HIPAA requirements with proper configuration.
Data Residency
On-premise deployment gives you complete control over data location.
Deployment
Cloud or on-premise. You decide.
ZenSearch Cloud
Managed hosting at app.zensearch.ai. We run the infrastructure; you run your team.
- Automatic updates
- Managed backups
- 99.9% uptime SLA (Enterprise)
- SOC 2 compliance (in progress)
On-Premise
On-premise AI deployment. Full control, full privacy, full customization.
- Docker & Kubernetes
- Air-gapped support
- Custom VPC / network
- Bring your own LLM
Security Review
Send us your
questionnaire.
We'll walk your team through the architecture and complete your security questionnaire.