Security

Your environment. Your data. Your call.

Ovatio deploys inside your environment and works with the systems you already run. Data and deployment stay under your control, not locked inside someone else's cloud. Every AI decision is logged where you can see it, and remediation only happens after a person approves it.

Security posture, product by product

Ovatio AI is built for evaluators who need clear answers: where policy is enforced, what gets logged, how access is checked, and which claims are supported by the current implementation.

Encryption
Access control
Approval gates
Audit trail
Observability

Ovatio Comply

  • MCP and REST proxy paths enforce policy before requests reach target systems and filter responses before they return.
  • Audit writes use persisted SHA-256 chain state with live hash-chain verification.
  • Bootstrap admin authentication with MFA is live; the current shipped admin model is intentionally narrow.
  • SQLite and PostgreSQL are supported runtime database modes.
  • OpenTelemetry wiring emits proxy, discovery, manifest, queue-depth, and health signals.

OvatioIQ Knowledge

  • ACL, RBAC, tenant-aware enforcement, and redaction are part of the security surface.
  • The Universal Plugin Framework encrypts secret configuration, verifies inbound webhooks with HMAC, and blocks SSRF-prone outbound destinations.
  • The recorded test catalog shows 1,213 tests passing, including security and MCP categories.
  • Air-gap deployment is supported through local Ollama inference, Qdrant, and SQLite.
  • Production deployment uses PostgreSQL and Qdrant.

OvatioIQ Agent

  • JWT authentication, bcrypt password hashing, token revocation, and login attempt lockouts are implemented.
  • Security audit logging records authentication and authorization events.
  • The MCP manager merges tools from multiple configured MCP servers into one agent tool set.
  • Known and configured system-of-record changes pause behind a human approval card before execution.
  • Read-only tool calls run in parallel where possible; read-only calls batched with a change are held until approval.

Important boundary

Agent approval is strongest when tools are known, configured, and risk-classified. For broader or dynamically extensible MCP deployments, pair Agent with Ovatio Comply so tool calls traverse a dedicated governance layer with policy enforcement, response filtering, and audit evidence.

Bring your security review.

We will map the claims to code, architecture, and operating evidence.