AI-Powered Testing Built to Enterprise Standards

Yes. Letaria uses cloud orchestration with on-premises execution agents. Deploy agents inside your corporate network, VPC, or even air-gapped environments. Letaria orchestrates jobs via encrypted connections, while all browser automation and test evidence stays within your infrastructure.

Every AI decision passes through our 6-layer guardrail system: G1 sanitises DOM input, G2 validates LLM output for dangerous patterns, G3 verifies selectors actually exist in the page DOM before execution (selector grounding), G4 checks reasoning-action consistency, G5 provides circuit breaker protection, and G6 logs a complete audit trail. This multi-layered approach catches hallucinated selectors, prompt injection attempts, and inconsistent reasoning.

Letaria aligns with NIST AI Risk Management Framework (AI RMF) for AI governance, SOC 2 Type II for security controls, ISO 27001 for information security management, and Hong Kong PDPO for personal data protection. Our audit trail, approval workflows, and explainable AI features are designed to satisfy regulatory review requirements across all industries and jurisdictions.

Yes. Enterprise deployments support Bring Your Own Keys (BYOK). You can use your own API keys for OpenAI, Anthropic Claude, Google Gemini, or other LLM providers. Your data is processed through your accounts and is never used for third-party model training. This gives you full control over data sovereignty and cost management.

Letaria supports policy-based approval workflows with four modes: any-one (single approver), all-required (every designated approver), sequential (ordered approval chain), and quorum (minimum number of approvals). Each policy can enforce comment requirements, specify required approvers, and filter by artifact type. All actions are logged in an immutable audit trail.

Yes. Enterprise deployments can be configured for Hong Kong data residency. Self-hosted execution agents keep test execution data within your HK infrastructure. Orchestration can be deployed in HK-region cloud providers. Storage for screenshots and reports uses tenant-isolated buckets with configurable retention policies.

Hong Kong is the #1 financial centre in APAC and #3 globally (after New York and London). It has world-class regulatory standards from bodies like the HKMA and SFC that set a high bar for enterprise software quality and compliance. We start here because Hong Kong provides abundant practical use cases and demanding requirements. Meeting these standards means Letaria already exceeds what most global enterprises need. We are actively expanding to optimise for enterprises in other APAC regions and beyond.

Letaria supports specialized execution agents for a wide range of enterprise platforms. The Web Browser Execution Agent (Playwright/Chromium) is available now. An API Execution Agent (REST/GraphQL) is coming soon. On our roadmap: Windows Desktop (.NET/WPF/WinForms), SAP GUI, Mainframe Terminals (3270/5250/AS400), Citrix/VDI, Mobile (iOS/Android via Appium), Java Desktop (Swing/AWT), and Oracle Forms. The same cloud orchestration drives all agent types — you can also request a custom execution agent for your specific platform.

Most AI testing tools use a single AI model to make decisions. Some newer tools are beginning to explore agentic features. Letaria is architecturally different: it deploys 6+ specialized AI agents — Step Interpreter, Element Locator, Input Guard, Output Validator, Consistency Checker, and Circuit Breaker — that cross-validate each other before any action is executed. This multi-agent validation pipeline catches hallucinations, inconsistencies, and dangerous patterns that a single model would miss.