Quayside Documentation
Quayside is a self-hosted control plane for the AI agents in your organisation. It sits between agents and the LLM providers they call, giving you:
- a registry of every agent class that exists
- an inline proxy that runs detection and policy checks on every call
- per-class policies with cascade, budgets, fail-mode, and on-failure handlers
- a complete audit trail of every call and every detector verdict
- a shadow log that surfaces unregistered agents the moment they try to call
This documentation is for the people who will deploy, operate, and integrate with Quayside.
Where to start
| You are… | Read |
|---|---|
| Trying it for the first time | Getting Started → Quickstart |
| Deciding whether Quayside fits your shape | Concepts → Architecture overview |
| Building an integration | Reference → HTTP API and Reference → Policy schema |
| Writing a custom detector | Guides → Writing a custom detector |
| Wiring CI/CD or scripts | Reference → CLI |
| Planning a deployment | Deployment → Self-hosting |
Sections
- Getting Started — install, quickstart, mental model
- Concepts — what the moving parts are and how they fit
- Guides — task-oriented how-tos
- Reference — HTTP API, policy schema, CLI, configuration
- Deployment — operating Quayside in production
Version
This documentation tracks Quayside 0.1.0 (pre-release).
License & support
Pre-release. License model not finalised. Issue tracking and customer-support channels will be documented here when they exist.