These docs describe Guardrail’s public interface: the policy format, the HTTP API, the SDKs, and the deployment setup. The engine itself is licensed source — it ships to licensees under a commercial license agreement, so request access to get it. All performance figures reference the reproducible benchmark.

• Quickstart — five minutes to a running Guardrail: the demo, your first policy, your first check.
• Threat Model — what Guardrail blocks, how it blocks it, and what it deliberately does not protect against.
• Policy Tuning — write and tune policies: categories, actions, param keys, bridge rules, session tracking, rate limits.
• Deployment — system requirements, building, Docker, systemd, nginx, TLS, health checks, load balancing.
• Integration — HTTP API reference, embedded library usage, and agent framework wiring (Claude, OpenAI, LangChain, MCP).
• SLI / SLO — service level indicators, SLO targets, monitoring queries, alert rules, error budgets.
• Benchmark — the reproducible performance and accuracy numbers, and how to run them yourself.
• Python SDK — install and use the Python SDK, MCP server, and inline proxy.