Documentation

Have natural conversations with AI about your ServiceNow environment. Ask questions about tables, records, scripts, and best practices. The assistant understands ServiceNow concepts and can provide contextual guidance based on your connected instance.

• Natural language queries about your instance
• Context-aware responses using your instance metadata
• Smart router across Anthropic, OpenAI (including Codex subscription), Gemini, OpenRouter, and self-hosted open-source models
• Streaming responses with real-time token and cost tracking
• Conversation history stored in Supabase with row-level security; accessible only to you

Beyond chat, Agent Mode lets Glidy AI execute multi-step operations on your ServiceNow instance - but never bypasses your approval. Risk-gated tool execution: READ operations run automatically; WRITE, DELETE, and ADMIN operations require explicit approval through the chat UI. Every action is visible, every write is reviewed.

• Multi-step task planning with explicit approval per write
• Risk-gated tools: READ auto, WRITE/DELETE/ADMIN explicit
• Update-set-aware writes - every change tracked in the right scope
• Live script generation, deployment, and runtime debugging
• Workflow and business rule creation with explicit user confirmation

Supercharge Mode enables per-instance code execution. When active, Glidy AI can write ES5 JavaScript directly and execute it on your ServiceNow instance's Rhino runtime - composing multi-record GlideRecord queries, transformations, joins, aggregations, and bulk writes into a single execution.

• Dramatically fewer LLM round trips on data-heavy tasks (audits, impact analyses, bulk migrations)
• Lower token cost; faster end-to-end completion
• Handles work that pure tool-calling agents handle slowly or not at all
• Must be explicitly enabled per instance (admin access required)
• All operations are logged and rate-limited

Connect and manage multiple ServiceNow instances from a single interface. Quickly switch between DEV, TEST, and PROD environments, or work across instances simultaneously with Cross-Instance Mode.

• Add unlimited instances (PDI, dev, test, production)
• Quick-switch between instances in the chat interface
• Cross-Instance Mode for operations spanning multiple environments
• Independent credentials and configuration per instance
• Connection testing to verify instance accessibility

Glidy AI uses a BYOK model. You provide your own LLM provider key - Anthropic, OpenAI (including the Codex subscription with subsidised tokens), Google Gemini, OpenRouter, or a self-hosted open-source model - giving you full control over costs, model selection, and usage. We never mark up AI costs.

• Smart router across Anthropic Claude, OpenAI (incl. Codex subscription), Gemini, OpenRouter, and self-hosted models
• 100+ models accessible via OpenRouter (Llama, Mistral, DeepSeek, Qwen, and more)
• Pay your AI provider directly at their published rates
• Real-time cost tracking per conversation
• Switch providers with a config change, not a vendor migration

Glidy AI is designed with security as a priority. ServiceNow credentials and LLM API keys are encrypted server-side; conversations are isolated per user with row-level security at the database layer; and we collect only minimal anonymized analytics.

• AES-256-GCM encryption for ServiceNow credentials and LLM API keys (server-side credential vault)
• OAuth 2.0 support with automatic token refresh
• Conversation history stored in Supabase with row-level security; encryption at rest and in transit
• Per-user data isolation enforced at the database layer
• GDPR-compliant data handling with self-service export and deletion
• Content Security Policy (CSP) and security headers enforced