Designed for DoW IL4/IL5/IL6 and FedRAMP High from day one. Every layer hardened, every connection monitored, every image signed.
Mutual TLS enforcement with DoW Common Access Card and PIV smart cards. Full certificate chain validation at the ingress layer.
Hardware-bound mutual TLS with YubiKey or any X.509 client certificate. Upload CA bundles directly from the admin UI with multi-cert support.
Azure Entra ID, Google Workspace, Okta, or any OIDC provider. Configurable allowed email domains. Works with Azure Gov sovereign clouds.
CIDR-based source IP filtering with real client IP preservation behind Azure Load Balancer. Combine with any other auth method.
Locked-down preset blocks all outbound traffic. Whitelist only the services your agent needs — AI providers, email, messaging, package registries.
Curated egress rules across 8 categories: AI providers, messaging, email, cloud services, gov cloud, certificate validation endpoints, and more.
Add custom domains and wildcard patterns. Layer 7 hostname filtering — not just IP-based rules.
Dedicated rules for AWS GovCloud, Azure Government, and federal endpoints. Toggle individually per tenant.
Continuous monitoring across multiple dimensions: processes, network activity, configuration state, filesystem integrity, resource usage, and more.
Critical findings trigger instant network isolation — containing threats before an operator even sees the alert. One-click restoration from the admin UI.
Aggregated view across all tenants. Severity filtering (Critical/High/Medium/Low), tenant status dots, CSV/JSON export for SIEM integration.
OHaaS records API activity, authentication events, and administrative actions to astructured audit log. We are evaluating immutable logging backends for production readiness.
Chainguard FIPS 140-3 validated cryptographic modules across all container images. No non-FIPS crypto anywhere in the stack.
Chainguard minimal base images with continuous CVE scanning on every build. CI/CD pipeline gates on High and Critical vulnerabilities — no exceptions.
Every image is cryptographically signed before publishing — no unsigned image ever reaches production. SBOMs generated and attached for full supply chain transparency.
OHaaS containers run as non-root with minimal attack surface. We areworking toward formal alignment with DISA STIG benchmarks.
Presidio-based PII and credential detection on all outbound traffic. Catches SSNs, credit card numbers, API keys, private keys, and JWTs before they leave your tenant.
44 regex patterns plus heuristic scoring on all inbound response bodies. Detects and blocks prompt injection attempts before they reach your AI agent.
11 YARA rules scan agent-generated commands in real time. Catches reverse shells, data exfiltration attempts, cryptominers, and container escape techniques.
Centralized management of all content security policies via the admin UI. View scan results, tune detection thresholds, and export findings for compliance reporting.
Each tenant gets dedicated compute, storage, network policies, and resource limits. Complete blast radius isolation — no shared resources between tenants.
Automatic mutual TLS between all services, Layer 7 traffic policies, and hostname-level outbound control. Zero-trust networking by default.
Root access disabled by default. Per-tenant toggle for package installation privileges — carefully scoped without container escape capabilities.
Export your entire AI agent as a single tarball — memory, skills, scripts, databases. Import into any OHaaS deployment or self-hosted instance.
OHaaS inspects every message in and out of your AI agent — detecting threats, stopping data leaks, and blocking malicious activity in real time.
Monitors every response from your AI model for attempts to hijack agent behavior — including jailbreak attempts, instruction overrides, and commands designed to exfiltrate data.
✓ Detects jailbreak and override attempts
✓ Configurable sensitivity: Low / Medium / High
✓ Log-only or block mode per tenant
Get started with OHaaS — the enterprise platform for running AI agents in regulated environments.