TuringTrust App
Four production releases since launch. One platform to route, secure, monitor, and govern every LLM call across your organization — live at app.turingtrust.ai
Start governing your AI stack today.
Free tier includes 5 users, 50 managed messages, and 200 BYOK messages per month. No credit card required. Full governance from day one.
FastAPI · React / Vite · PostgreSQL · MIT-licensed SDK
Built for the teams who own AI risk.
From real-time monitoring to one-click compliance exports — every screen is designed for data governors, CISOs, and platform engineers.
Everything in one place.
Nothing bolted on.
TuringTrust ships as a single integrated platform — not a collection of point solutions. Every module shares the same policy engine, audit log, and data model.
Governance & Control
Unified LLM Gateway
Single proxy for OpenAI, Anthropic, Google/Gemini, Groq, and DeepSeek. Encrypted key vault (AES-256-GCM), multi-key fallback, circuit breaker, and per-provider timeout controls.
Policy Engine
20 deterministic operators with in-memory evaluation (<1ms). Five governance actions: ALLOW, BLOCK, REDACT, WARN, LOG_ONLY. Version-controlled policies with full rollback support.
Approval Workflows
Multi-step queues with configurable risk thresholds. Auto-approve below threshold, require human sign-off above. Full decision audit trail and webhook integrations.
Team Management & RBAC
Six roles: super_admin, org_admin, team_admin, admin, approver, user. Department-level model access controls. SSO/SAML 2.0 for Compliance+ tiers.
AI Security & PII
Two-Tier REDACT Pipeline
Tier 1: fast regex + NER scan. Tier 2: context-aware LLM review for ambiguous cases. Sub-50ms latency, near-zero false negatives. Catches what rule-based systems miss.
Red Team Console
Pre-loaded adversarial library: 20+ prompt injection vectors, 10+ jailbreak techniques, PII extraction, RAG poisoning, bias stress tests. CWE/OWASP mapping with risk severity scoring.
Agent Guardian
Runtime security for autonomous AI agents: velocity limits, domain whitelists/blacklists, cost caps (max USD/hour), emergency stop webhooks. Three modes: audit, enforce, escalate.
Anomaly Detection
Seven statistical algorithms running continuously: Z-score token spikes, unusual model use, off-hours access, cost surges, error rate spikes (>20%/hr), rapid fire (>10 req/min). Auto-severity classification.
Compliance & FinOps
Compliance Certifier & Audit
850+ controls mapped to EU AI Act, HIPAA, SOC 2, GDPR, India DPDP Act, and ISO 42001. One-click PDF/JSON evidence packs for external auditors. Immutable logs — no DELETE or PUT routes exist in the API by design. Live posture scoring per framework. SIEM integrations with Splunk, Datadog, and ELK.
FinOps & Token Tracking
Per-model cost breakdowns by team, user, and department. Budget controls with alerts and monthly resets. BYOK (zero markup) or TuringTrust Managed (35% fee, 10–20% cheaper than Bedrock/Azure). Forecasting and optimization insights.
Model Arena & Benchmarking
Upload golden datasets (CSV), run identical prompts across all providers simultaneously. Compare cost-per-success, hallucination rate, latency, and token counts. Leaderboard with side-by-side results for data-driven model selection.
Shipping fast.
Four versions. Three months.
Every release of “Sentinel” adds production-grade capabilities. No feature flags, no beta caveats — shipped means shipped.
June 2026
v1.3.0 “Sentinel”
- + India DPDP Act & ISO 42001 evidence packs
- + SSO / SAML 2.0 enterprise identity integration
- + DeepSeek V3 provider support
- + Live compliance posture scoring per framework
- + Member Activity analytics with department-level drill-down
v1.2.0
May 2026
v1.2.0 “Sentinel”
- + Agent Guardian — runtime security for autonomous AI agents
- + Red Team Console with CWE/OWASP mapping
- + Compliance Certifier: 850+ controls, EU AI Act evidence packs
- + SIEM integrations: Splunk, Datadog, ELK
- + Approval workflows with configurable risk thresholds
v1.1.0
April 2026
v1.1.0 “Sentinel”
- + Tier 2 LLM-based PII detection — context-aware NER
- + Anomaly Detection — 7 statistical algorithms, auto-severity
- + Model Arena & Benchmarking (golden dataset upload)
- + Budget controls with monthly resets and email alerts
- + Multi-key fallback per provider with background health monitoring
v1.0.0
March 2026
v1.0.0 “Sentinel” — Initial Launch
- + Unified LLM Gateway (OpenAI, Anthropic, Google, Groq)
- + REDACT Pipeline — Tier 1 regex + NER PII detection
- + Policy Engine with 20 operators and 5 governance actions
- + FinOps dashboard with per-model token tracking
- + 6-role RBAC, multi-tenant RLS, immutable audit logs
- + AI Chat interface with BYOK and Managed key modes
Production-grade from day one.
Built on a stack trusted by enterprises. The SDK is MIT-licensed — no secrets, no policy logic in your codebase.
FastAPI Backend
62 backend services. Async Python with PostgreSQL and Row-Level Security for multi-tenant data isolation. Per-provider timeouts with circuit breaker patterns.
FastAPI · PostgreSQL · RLSReact / Vite Frontend
42 frontend pages. Type-safe, fast-refresh development. Real-time streaming with Server-Sent Events. Accessible, keyboard-navigable interface.
React · Vite · TypeScriptMIT Python SDK
~25KB thin connector. All governance runs on TuringTrust servers — nothing sensitive ships in your codebase. Drop-in replacement for OpenAI, Anthropic, Google, and Groq clients.
pip install turingtrustThe platform is live.
Your turn.
Sign up for free and have your first governed LLM call running in under five minutes. No credit card, no commitment.
Or explore the pricing plans to find the right tier for your team.