Amurg Documentation

Amurg is a self-hosted control plane for deploying, managing, and interacting with AI agents through a mobile-friendly chat UI. It consists of three components — a Hub that handles authentication, routing, and persistence; a Runtime that spawns and manages agent processes; and a UI that provides the browser-based chat interface.

This documentation covers everything you need to configure, deploy, and extend Amurg. Choose a topic below to get started.

Configuration

Hub and runtime JSON config reference. Server settings, auth, storage, session behavior, and endpoint definitions.

Adapters

All 8 adapter profiles with config fields, capability tables, and example JSON for each.

API Reference

HTTP endpoints, request and response schemas, authentication headers, and error codes.

WebSocket Protocol

Real-time message types, envelope format, session lifecycle, and bidirectional flows.

Voice Input

Voice dictation with browser Web Speech API or a local Whisper server. Hold-to-talk, tap-to-toggle, edit before send.

Deployment

Docker Compose setup, production configuration, reverse proxy, TLS, and monitoring.

Security

Authentication providers, JWT tokens, endpoint authorization, permission modes, and audit logging.

Getting Started

If this is your first time setting up Amurg, start with the Configuration page. It walks through both the hub and runtime config files, including how to define endpoints and connect a runtime to your hub instance.

For a quick local setup: build with make build, start the hub with the local config, start a runtime, and open the UI on port 3000. Default credentials are admin / admin.