Products
Sovraine Guard evaluates.
Sovraine One commands.
One free governance engine for every agent decision. One enterprise platform to operate your entire AI fleet.
SOVRAINE GUARD
Evaluate
The cascade decides in milliseconds. Policy-as-code, fail-closed, audited.
SOVRAINE ONE
Deliberate
Multiple agents debate. Ops, Security, Compliance, Legal — each brings their perspective.
TOGETHER
Govern
Every AI decision — evaluated, debated, approved, executed, and proven. End to end.
Sovraine Guard
The governance engine. Evaluates every AI agent action before execution — in 12ms, with a cryptographic proof.
6-Stage Cascade
Behavioral detection, agent constraints, cache, policies, debate, human escalation.
Policy-as-Code
.guard.md Markdown with YAML frontmatter. Version controlled, auditable.
Community Hub
650+ verbs, 104 MCP server mappings, 2,100+ tools, 18 sectors. Hub content under CC-BY-SA-4.0.
Immutable Audit
SHA-256 hash chain, Ed25519 signing, JSONL export. Tamper-evident.
# Initialize with sector policies
sovctl guard init --sector financial-services
# Evaluate an action
sovctl guard eval --verb reverse-payment \
--target production/payments-api \
--context environment=production amount=125000
# Compliance report
sovctl guard compliance --framework eu-ai-act
Sovraine One
Before any AI agent acts on your infrastructure, Sovraine One convenes the right stakeholders — Ops, Legal, Compliance, Finance — and forces a verdict. In milliseconds. With a signed record.
Debate Engine
Ops, Security, Compliance, Legal, Finance agents. Risk-routed. Budget-controlled.
RBAC & Identity
Keycloak OIDC, 4 roles (admin, viewer, sector-admin, auditor).
Interop Connectors
Bidirectional sync with OPA, Cedar, K8s Admission, ServiceNow, Splunk.
Policy OTA
Over-the-air policy updates, custom scanner plugins, audit-ready compliance evidence export.
In active development — reserve early access and we'll reach out as launch nears.
Solutions
One platform. Four problems it solves.
However you frame the risk — rogue agents, exposed tools, audit gaps, or scale — Sovraine is the layer between intent and execution.
AI Agent Governance
Give every agent its own risk ceiling, skills, and constraints. Each action runs the 6-stage cascade and returns one of six verdicts — ALLOW, DENY, WARN, FILTER, ESCALATE, HUMAN — fail-closed by default.
- ✓ Per-agent constraints & risk ceilings
- ✓ Two checkpoints baked into the agent loop
- ✓ Policy-as-code in version-controlled .guard.md
MCP Security
Wrap any MCP server with zero code changes. Every tool call is intercepted, classified, and audited before it executes — and sensitive data is redacted, routed locally, then rehydrated by the privacy router.
- ✓ 6 certification gates per server
- ✓ 2,100+ tools classified across 104 servers
- ✓ PII redact / local-route / rehydrate
Compliance Automation
Every decision is logged in a SHA-256 hash chain and mapped to the frameworks your auditors ask about — exportable on demand, not reconstructed after the fact.
- ✓ NIST AI RMF, EU AI Act, ISO 27001, OWASP LLM Top 10, MITRE ATLAS, SOC 2, HIPAA, GDPR
- ✓ Immutable, tamper-evident audit trail
- ✓ Enterprise adds continuous control evidence matching
Enterprise Deployment
Run Sovraine One on-premise or in your private cloud with full fleet control — role-based access, identity, interop with your existing policy stack, and over-the-air policy updates.
- ✓ RBAC + Keycloak OIDC identity
- ✓ Interop: OPA, Cedar, K8s Admission, ServiceNow, Splunk
- ✓ On-premise & private cloud, policy OTA