Home
A local bridge between IFC models and AI tools. ifc-console loads a model
with IfcOpenShell and exposes it through MCP, Python, an interactive terminal,
and optional browser tools.
+------------+ MCP (streamable HTTP / stdio) +---------------------------+
| LLM client | --------------------------------> | ifc-console (one process) |
| Claude ... | | +- MCP tool layer |
+------------+ | +- IfcOpenShell session |
| +- Policy engine (modes) |
you, in a terminal --------------------> | +- Interactive console |
your browser (optional 3D viewer) ------> | +- Web viewer (localhost)|
+---------------------------+
Start here¶
uv tool install ifc-console
cd path/to/your/models
ifc-console
In the console, run /file, then /connect <client>. The setup is one-time;
after that, choose a model and ask questions from your AI client.
> /file
> /connect codex
> /viewer
Follow the first-session guide
Where next¶
| goal | page |
|---|---|
| learn the terminal | Console |
| connect Claude, Cursor, VS Code, or Codex | Connecting clients |
| understand editing and saving | Safety |
| use the browser tools | 3D viewer and Chat |
| automate with Python | Python SDK |
| build an agent | Agent applications |
| run repeatable checks | Workflows |
| look up commands | CLI and MCP tools |
| fix a problem | Troubleshooting |
The base install contains no browser assets. Add ifc-console[viewer] only
when you need the local 3D viewer and chat panel.