# OptScale AI — AI Governance Platform for the Enterprise
 
> OptScale AI is an enterprise AI governance platform developed by Hystax that unifies local LLMs and public AI APIs through a single, governed control plane. It provides intelligent routing, real-time security guardrails, full observability, AI agent governance, AI cost optimization with token compression, and business-outcome ROI measurement — all deployable as cloud SaaS or on-premises. Built on a decade of cloud FinOps expertise, OptScale AI brings the same cost discipline to every LLM call and MCP interaction. It is OpenAI API-compatible — teams switch their base URL and all existing code works without changes.
 
OptScale AI is the product of Hystax, a company founded in 2016 on cloud cost optimization (FinOps) and trusted by enterprises including Airbus, Nokia, T-Systems, DHL, and PwC. The platform is available as a cloud-hosted SaaS offering or as an air-gapped on-premises installation for regulated industries with data sovereignty requirements.
 
The platform is organized around six core pillars: an Intelligent AI Gateway, AI Security & Guardrails, Team & Agent AI Performance, AI Agent Control, AI Cost Optimization & Token Compression, and AI ROI & Impact. Its core promise: cut AI costs by up to 60% while governing every prompt.
 
## Key Performance Characteristics
 
- Average LLM cost reduction: up to 60% (via smart routing, model arbitrage, and token compression)
- Token billing reduction: up to 97% fewer tokens billed at the full rate
- Gateway routing / compression overhead: <10ms added latency
- Gateway uptime SLA: 99.9% (up to 99.99% on Enterprise)
- Request visibility: 100%, with no token masking on prompts and responses
- Audit log retention: 90 days (default), custom retention available
- MCP connectors: 20+ pre-built, new connectors added monthly
 
## Core Product
 
- [Home — OptScale AI: AI Governance Platform for the Enterprise](https://optscale.ai/): Overview of the platform, key value propositions, six pillars of AI governance, deployment options, and integration ecosystem.
 
## Product Pillars
 
- [AI Gateway](https://optscale.ai/product/ai-gateway): Intelligent routing of queries to the optimal model or MCP server based on task complexity, cost, and latency. Includes prompt compression, model output arbitrage, role-based access control, full audit logging, data sovereignty controls, and drop-in OpenAI API compatibility. Average cost reduction of up to 60% through smart routing and local LLM offloading.
 
- [AI Security & Guardrails](https://optscale.ai/product/ai-security-and-guardrails): Real-time content filtering, PII detection and redaction (50+ entity types), data loss prevention (DLP), and policy enforcement at the gateway level — before prompts reach any external model. GDPR, HIPAA, and PCI-DSS aware. Policies are defined in YAML, versioned, and scoped per team, model, or company-wide.
 
- [Team & Agent AI Performance](https://optscale.ai/product/model-analytics-and-tracing): Usage analytics broken down by team and agent, efficiency scores and leaderboards ranking output value per dollar, side-by-side model benchmarking on quality/latency/cost, full prompt visibility with no token masking, end-to-end distributed tracing (P50/P95/P99 latency), and configurable audit-ready log retention. Positioning: "you can't govern what you can't see."
 
- [AI Agent Control](https://optscale.ai/product/ai-agent-control): Governance layer for AI agents your teams have already built. OptScale AI does not run agents — it governs them. Register agents built with LangChain, CrewAI, AutoGen, or custom code. Set hard cost limits, time limits, and recursion depth per agent. Whitelist exactly which MCP servers and vector stores each agent can reach — unauthorized access attempts are blocked and logged. Real-time anomaly detection catches loops, token bursts, and behavioral drift before they escalate. Every action is logged — who, what, when, why — compliance-ready and replayable.
 
- [AI Cost Optimization & Token Compression](https://optscale.ai/product/ai-cost-optimization-token-compression): Four stacked optimization mechanisms — system-prompt & schema trimming, a proprietary lossless inline compression engine, LLM cache alignment (prefixes charged at ~10% of standard rates), and smart model selection — that cut LLM spend by up to 60% and reduce fully-billed tokens by up to 97% with <10ms added latency and no application code changes. Typical reductions: tool outputs/logs and code/search results up to ~90%, RAG chunks 40–75%, conversation history 50–80%.
 
- [AI ROI & Impact](https://optscale.ai/product/ai-roi-and-impact): Connects AI infrastructure cost to actual business outcomes by joining gateway telemetry with systems of record (GitHub, Jira) on trace_id, commit, and pull request. Delivers derived-not-declared metrics such as cost per merged PR, cost per feature, revert rates, spend by team/model, and AI-assisted merge percentage. Role-based metric packs for Engineering, Finance, and Adoption. Individual-level analytics ship disabled by default with audit logging and small-group suppression; requires no gateway replacement.
 
## Solutions
 
### By Industry
 
- [Banking & Financial Services](https://optscale.ai/solutions/banking-financial-services): AI governance for lending, fraud, trading, and advisory teams — compliance, audit trails, and data protection built for regulated financial institutions.
 
- [Media & Communication](https://optscale.ai/solutions/media-communication): Governed AI adoption for media and communication companies — content workflows, cost control for high-volume AI usage, and brand-safe guardrails.
 
- [Education](https://optscale.ai/solutions/education): AI governance for schools and universities — student data boundaries, FERPA and COPPA compliance, and a full audit trail across classroom and administrative AI use.
 
- [Technology & Software](https://optscale.ai/solutions/technology-software): AI governance for technology and software companies shipping AI-powered products — one endpoint for every model, guardrail, and agent.
 
### By Use Case & Role
 
- [Finance & Accounting](https://optscale.ai/solutions/finance-accounting): AI cost visibility, department-level billing, and the audit evidence finance and accounting teams need.
 
- [Security & Compliance](https://optscale.ai/solutions/security-compliance): Centralized guardrails, DLP, and agent governance for security and compliance teams responsible for enterprise AI usage.
 
- [Engineering Teams](https://optscale.ai/solutions/for-engineers/): Ship AI products without building governance from scratch — drop-in gateway, observability, and agent controls for engineering and platform teams.
 
## Pricing & Plans
 
### SaaS (Cloud) — per seat per month, requests pool across the team
 
- Free: $0 — up to 5 seats, AI Gateway with smart routing, basic guardrails, cost dashboard, 5 MCP connectors, 7-day log retention, community support
- Starter: $12/seat/month — min 5 seats, advanced filtering, PII detection, basic tracing, role-based access control, 10 MCP connectors, 30-day retention, email support (99.5% SLA)
- Business: $29/seat/month — min 10 seats, prompt optimization, model arbitrage, DLP, full tracing & debugging, RAG, department-level billing, 100 MCP connectors, 90-day retention, priority support (99.9% SLA)
- Enterprise: custom pricing — unlimited seats and requests, SSO/SAML, custom retention, audit export, dedicated success manager (99.99% SLA), white-glove onboarding
 
### On-Premises — annual license by cluster size, unlimited users and requests
 
- Small Cluster: $6,500/year — up to 5 nodes, 40 vCPUs, all 20+ MCP connectors, up to 10 agents, up to 3 local models, business-hours support
- Medium Cluster: $32,000/year — 6–20 nodes, 160 vCPUs, GPU support, up to 50 agents, up to 10 local models, air-gapped installation, 24/7 priority support
- Large Cluster: $85,000+/year — 21+ nodes, custom vCPUs, unlimited agents and models, custom MCP connectors, dedicated engineer, white-glove onboarding
 
- [Full Pricing Page](https://optscale.ai/pricing)
 
## Integrations & MCP Connectors
 
Pre-built MCP server connectors across categories:
 
- Sales & CRM: Salesforce, HubSpot, Pipedrive
- Engineering: GitHub, Jira, AWS, ServiceNow
- Finance & Ops: QuickBooks, Stripe
- Collaboration: Slack, Confluence, Google Docs, Notion
- Identity & Security: Okta, Azure AD
- Custom: MCP Server SDK, REST API Adapter, Webhook Connector
 
New connectors ship monthly. Enterprise customers can request priority connectors.
 
## Documentation
 
- [Documentation Home](https://optscale.ai/documentation/): Technical documentation for OptScale AI — setup guides, API references, integration instructions, and configuration details for all platform pillars.
- [Architecture Overview](https://optscale.ai/documentation/architecture-overview/): Platform architecture — gateway, policy engine, tracing, and agent control plane.
- [First Steps](https://optscale.ai/documentation/first-steps/): Getting started with OptScale AI — account setup and first requests through the gateway.
- [Switch to the Gateway](https://optscale.ai/documentation/switch-to-gateway/): How to point existing OpenAI-compatible code at the OptScale AI gateway by changing the base URL.
 
### Administration — Core Services
 
- [Core Services Overview](https://optscale.ai/documentation/admin/core-services/)
- [Dashboards](https://optscale.ai/documentation/admin/core-services/dashboards/)
- [AI Access](https://optscale.ai/documentation/admin/core-services/ai-access/)
- [Providers](https://optscale.ai/documentation/admin/core-services/providers/)
- [Policies & Guardrails](https://optscale.ai/documentation/admin/core-services/policies-guardrails/)
- [MCP Servers](https://optscale.ai/documentation/admin/core-services/mcp-servers/)
- [Usage & Cost](https://optscale.ai/documentation/admin/core-services/usage-and-cost/)
- [Traces](https://optscale.ai/documentation/admin/core-services/traces/)
 
### Administration — Platform
 
- [Model Training](https://optscale.ai/documentation/admin/model-training/)
- [Experiment](https://optscale.ai/documentation/admin/experiment/)
- [Environments and Operations](https://optscale.ai/documentation/admin/environments-and-operations/)
- [Use Cases](https://optscale.ai/documentation/admin/use-cases/)
- [Limitations](https://optscale.ai/documentation/admin/limitations/)
- [End-to-End Example](https://optscale.ai/documentation/admin/end-to-end-example/)
 
### Chat
 
- [Interface Overview](https://optscale.ai/documentation/chat/interface-overview/)
- [Chats and Projects](https://optscale.ai/documentation/chat/chats-and-projects/)
- [Chat Use Cases](https://optscale.ai/documentation/chat/use-cases/)
 
### External Tools
 
- [External Tools Overview](https://optscale.ai/documentation/external-tools/)
- [OpenCode](https://optscale.ai/documentation/external-tools/opencode/)
- [Cursor](https://optscale.ai/documentation/external-tools/cursor/)
- [Claude](https://optscale.ai/documentation/external-tools/claude/)
 
## Blog
 
- [LLM Cost Attribution: Splitting AI Spend Fairly Across Teams, Departments & Agents](https://optscale.ai/blog/llm-cost-attribution-teams-departments-agents): How virtual keys make every model call traceable to a team, a department, or a single agent.
- [Runaway Agents: Catching Loops, Drift & Recursion](https://optscale.ai/blog/runaway-agents-loops-drift-recursion): The agent failure modes that quietly burn budget, and the observability metrics that catch them before they escalate.
- [Shadow AI Agents & the AI Agent Governance Gap](https://optscale.ai/blog/shadow-ai-agents): How unauthorized, internally built AI agents evade existing governance controls, and where oversight must begin.
- [Blog Home](https://optscale.ai/blog): All OptScale AI articles on AI governance, cost optimization, and agent observability.
 
## Partners
 
- [Partner Program](https://optscale.ai/partner-program): Reseller and technology partner program for OptScale AI.
 
## Company
 
- [About Hystax — The Company Behind OptScale](https://optscale.ai/company): History of Hystax from cloud migration (Acura platform) and FinOps (OptScale open-source) to AI governance. Enterprise customer list, deployment philosophy, and company background.
 
## Contact & Support
 
- [Contact](https://optscale.ai/contact-us): Sales inquiries, pricing questions, technical guidance, and on-premises deployment requests. Includes a live demo environment link.
 
- [Live Demo](https://my.optscale.ai/live-demo): Interactive live demo environment for OptScale AI. Explore the platform without installation.
 
- [Login](https://my.optscale.ai/login): Access to the OptScale AI cloud-hosted SaaS environment.
 
- [Sign Up (Free)](https://my.optscale.ai/register): Create a free OptScale AI account — up to 5 seats, no credit card required.
 
## Optional
 
- [Privacy Policy](https://optscale.ai/privacy-policy): Data collection, cookie usage, third-party sharing (Google Analytics, Facebook Pixel), and data transfer practices for website visitors and content downloaders. Governed by California law.
 
- [Terms of Use](https://optscale.ai/terms-of-use): End-user license agreement (EULA) for OptScale AI software. Covers usage rights, restrictions, warranty, liability limitations, and governing jurisdiction (California).