Welcome

The Aetherium Engine is a multi-agent AI orchestration platform that runs on your own hardware. You describe what you want built; a team of AI agents — a supervisor that plans and delegates, workers that write code, run commands, and check each other's output — collaborates to build it, inside sandboxed containers, on infrastructure you control. No cloud dependency, no per-token bill, no data leaving the box.

This is the Developer Documentation: the complete reference for understanding, running, extending, and operating the Engine. It is written for someone who evaluates a platform by reading it, not by watching a demo — so it favors specifics over pitch, names real services and real trade-offs, and tells you where the Engine knowingly stops short.

Start where you are

You don't need to read this front to back. Pick the path that matches what you're here to do.

Evaluating it. Start with §01 — What is the Engine for the premise and what makes it different, then §03 — System Architecture for how the pieces fit, then §17 — Architecture Decisions for the reasoning and the honest trade-offs behind the design. That's the arc for deciding whether the Engine's values match yours.

Getting hands-on. Go straight to §02 — Getting Started. It takes you from a clean clone to a green health check and a working multi-agent build — your first team writing real code — in one sitting.

Building on it. If you're adding capability, §10 — Development Workflow covers the day-to-day loop, §12 — Extending the Engine covers the extension points (new tools, MCP server classes, templates, skills), and §19 — Contributing covers the conventions and disciplines that keep the codebase sound.

Running it. If you're deploying and operating, §14 — Deployment gets the Engine onto a Linux host, §15 — Operations covers keeping it healthy, and §16 — Troubleshooting is the symptom-to-fix reference for when something breaks.

Hit an unfamiliar term anywhere? §20 — Glossary defines every piece of Aetherium-specific vocabulary, each linked back to the section that explains it in full.

The documentation at a glance

The twenty-one sections group into four arcs:

Orientation (§00–§04) — what the Engine is and the concepts it's built on: this welcome, What is the Engine, Getting Started, System Architecture, and Core Concepts.

Building (§05–§12) — the codebase and how to work in it: Solution Layout, Key Services, MCP Architecture, Templates, Configuration, Development Workflow, Benchmarking, and Extending the Engine.

Running (§13–§16) — taking it to production: Integrations, Deployment, Operations, and Troubleshooting.

Reasoning & reference (§17–§20) — the why and the lookups: Architecture Decisions, Roadmap, Contributing, and the Glossary.

What to expect

The documentation is direct about what the Engine does well and where it asks something of you. It runs locally and keeps your work on your hardware; it coordinates real multi-agent teams against real tasks; and it gates every tool call through a permission model built for code that runs shells and writes files. It also has rough edges and open trade-offs, named plainly where you'll meet them and gathered in §18 — Roadmap. You'll find real ports, real service names, and real configuration keys throughout — and the occasional honest "this is hardcoded today" or "this is the operator's job." That's the point: enough detail to actually run the thing, not a tour.

Pick the path that fits and start there.