Odock Documentation

Govern LLM and MCP traffic from one gateway.

Odock sits between applications, agents, developer tools, AI providers, and MCP servers so every model call and tool call can be authenticated, routed, checked, recorded, and monitored.

What is Odock?Quick start

Choose your path

Start with the core concepts, then move into usage, management, security, plugins, observability, or self-hosting.

Understand Odock

Learn where the AI governance gateway fits between applications, agents, providers, and MCP servers.

Make a first call

Configure Odock and send an OpenAI-compatible gateway request with a virtual API key.

Use the gateway

Call native provider-compatible endpoints or the unified Odock endpoint for multi-provider routing.

Configure Models & MCP

Manage providers, provider keys, models, endpoints, MCP servers, grants, pricing, and usage records.

Manage runtime controls

Set virtual API keys, routing policies, budgets, quotas, reservations, and usage boundaries.

Apply security guardrails

Layer RBAC, access grants, policies, SafetySec checks, plugin hooks, budgets, quotas, and telemetry.

Extend with plugins

Add request and response processing through lifecycle phases, marketplace plugins, or custom plugins.

Observe production traffic

Track usage monitoring, logs, traces, metrics, dashboards, alerts, and OpenTelemetry integration.

Self-host Odock

Run Postgres, Redis, Traefik, server, UI, migrations, and the optional observability stack with Docker Compose.