Skip to content

Scutum docs#

The self-hosted LLM gateway. One OpenAI-compatible endpoint, 100+ models across 9 providers, with cost governance, audit, routing, and policy built in. Postgres as source of truth. Your data never leaves your infra.

QuickstartGet a local Scutum stack running in under 5 minutes.

API integrationDrop-in OpenAI-compatible endpoint. Python, TypeScript, Go, curl.

Production readinessSizing, HA, security posture before you ship.

How it fits together#

Your apps ──► Scutum ──► OpenAI · Anthropic · Google · Mistral · xAI · DeepSeek · Bedrock · Azure · Vertex · Ollama
                │
                ├─ Cost governance      budgets, chargeback, predict-before-spend
                ├─ Audit & compliance   immutable trail, retention policies, exports
                ├─ Policy & guardrails  Cedar policies, DLP, prompt registry
                ├─ Routing & failover   model groups, fallback chains, A/B tests
                ├─ MCP & A2A            agent gateway with allowlists
                ├─ Observability        OpenTelemetry, Prometheus, Jaeger, Grafana
                ├─ Workflows            LangGraph templates, Temporal-backed agents
                └─ SRE agent            LLM-driven incident remediation, human-in-loop

What lives where#

  • Getting Started — install, point your SDK at it, see how it compares to alternatives.
  • Reference — feature deep-dives: routing, caching, guardrails, MCP, workflows, the SRE agent, cost management, the chat product.
  • Operations — admin console, licensing, observability, cloud deploys; production readiness, sizing, HA; security threat model + secret rotation; FinOps KPIs.
  • Reading — whitepapers and research notes for buyers and architects who want the longer-form rationale.

Need a hand? Reach us at [email protected] or book a demo.