Home AJ AI Brain
AJ AI Brain

Then I taught it what good work looks like.

A general AI model knows a great deal about everything and nothing about how your project names a duct. So the agent reads a proven knowledge package before it acts — task skills, techniques and working code captured from real projects — and then carries out real modelling work on the live, open Revit model through a Node.js MCP bridge.

What it is made of

9task workflows, from air terminal layout to NFPA 13 sprinkler spacing checks
267reusable Revit API C# fragments, compile-checked against the real API
23knowledge documents the agent reads before it acts
  • Team-wide, not personal. Packaged with an installer, version-controlled in Git, documented. The time saving is shared.
  • Quality built in. Automated checks compile-verify the C# against the real Revit API and flag documentation drift on every edit.
  • Built the way it runs. AJ-Tools itself is written with AI coding tools — Claude Code, Codex and Antigravity — against the same compile-checked codebase the agent draws from.
  • Honest tooling. A tool that has never been run against a real model says so.

The model is still the job. The tools just stop it eating the day.

Looking for an MEP BIM Coordinator?

Doha, Qatar · transferable QID with NOC · able to join within days.