Everything You Need to Govern AI
A complete platform that covers security, privacy, compliance, cost control, quality, and auditability.
Security Guardrails
Automated security filters that inspect every AI conversation on input AND output. 5 detection engines, 8 predefined rulepacks, custom YAML rules, AI-generated patterns, and real-time threat blocking.
Capabilities
- Basic Guardrails (local, 30+ regex patterns)
- NeMo Guardrails (NVIDIA ML-powered)
- Lakera Guard, Azure Content Safety, OpenAI Moderation
- Block or Warn actions per category
- Prompt injection & data exfiltration detection
- Custom rulepacks via YAML or visual dashboard
- AI-powered pattern generation from natural language
PII Detection & Redaction
Real-time scanning of every request and response for personally identifiable information. Automatic masking ensures the AI model never sees sensitive data, with smart rehydration to restore tokens in the response.
Capabilities
- Basic PII (Email, Phone, SSN, Credit Card, IP, IBAN)
- Presidio ML engine (Microsoft) with custom recognizers
- 8 industry-specific PII rulepacks
- Automatic redaction: [EMAIL], [SSN], [CREDIT_CARD]
- PII Rehydration — model never sees real data
- Risk levels: High, Medium, Low
- Per-customer rulepack assignment
Compliance Engine
Real-time regulatory evaluation powered by Open Policy Agent. Every request is assessed against EU AI Act and NIST AI RMF frameworks with automatic risk classification and violation tracking.
Capabilities
- EU AI Act (risk classification, transparency, human oversight)
- NIST AI RMF (Safe, Fair, Explainable, Reliable, Accountable)
- Multi-framework orchestrator evaluation
- Dynamic risk classification per prompt + model
- Use-case extraction (healthcare, finance, legal)
- Capability detection (emotion recognition, deepfakes, social scoring)
- Per-tenant framework selection and priority
Cost Analytics & FinOps
Complete cost tracking, visualization, and enforcement. Budget and rate limits verified BEFORE calling the LLM — not after. Per-provider and per-model granularity with overage billing integration.
Capabilities
- Budget limits per provider and model (daily/monthly)
- Rate limits per provider and model (RPM)
- Pre-execution enforcement (HTTP 429)
- 80% usage warning headers
- Cost trends by provider, model, time period
- Quota management with overage billing (Polar)
- Automatic window rotation
Quality & Performance Monitoring
End-to-end reliability monitoring of your AI infrastructure. Track latency, success rates, SLA compliance, and error trends with per-model performance grids.
Capabilities
- Latency metrics: avg, P95, P99 per model
- Success/error rates per model
- Configurable SLA targets with alerting
- Error trend visualization
- Model performance health grid
- Arize, WhyLabs, Evidently adapters (built)
Audit Trail
Immutable, chronological record of every significant event. 12 event types covering access, violations, redactions, limits, routing, and configuration changes — with Langfuse trace integration.
Capabilities
- 12 event types (model.access, compliance.violation, pii.redacted, ...)
- Filterable log with date range and event type
- CSV export of filtered audit logs
- Langfuse trace lookup by Trace ID
- Key indicators: total events, top event type, top actor
- Expandable event detail with full JSON payload
- Data minimization: no raw prompts or responses stored
Multi-Provider Routing
Intelligent multi-provider routing with health monitoring and automatic failover. 3 providers live in production with 7 more ready to connect.
Capabilities
- OpenAI, Anthropic, Ollama (live)
- HuggingFace, Vertex AI, Bedrock, Cohere, Mistral, vLLM, TGI (built)
- Weighted routing strategy
- Automatic fallback with incident logging
- SLO-aware and Multi-Armed Bandit strategies (built)
- 3-step provider wizard: credentials → models → limits
- Real-time health monitoring per provider
Enterprise Security & Access
Complete authentication and authorization built on Better Auth with zero-trust architecture. Multi-tenant isolation, RBAC, and full organization management.
Capabilities
- OAuth (Google, GitHub) + email/password + 2FA
- RBAC: Admin, Member, Viewer, Owner + custom roles
- Granular CRUD permissions per module
- Team and organization management
- API key management (create, rotate, revoke)
- Zero-trust mTLS with SPIFFE/SPIRE
- WORM audit logs, GDPR controls