An AI agent can now read your CAD and BIM models and answer in plain language.
ODA MCP Servers arrive September 2026.
You have a model open and a plain question about it. How many of this fitting are in the assembly. Whether the layout still matches last week's revision. Whether a corridor clears code. Answering has always meant opening the right software yourself, or waiting on someone who can write code against it. Now you can ask an AI agent, and it answers from the model.
That is what ODA MCP Servers do, starting September 2026. An MCP server is the connection between an AI agent and your engineering files. ODA's works with any MCP-capable agent and reads DWG, STEP and IFC at launch, with DGN, Revit and Navisworks to follow. It comes with the Core Package+ — ODA membership from the Sustaining tier up — royalty-free and licensed per company, not per seat. You run it on your own hardware.
What changes is who can do this work. For 28 years, using ODA meant writing code against an SDK — the reach of a developer. Now the person holding the question can get to the answer without that step. As ODA president Neil Peterson puts it:
“This will put builder-like capabilities into the hands of a much larger audience, well beyond our current developer audience.”
The trade press treated this as a real shift, not a feature. In AEC Magazine, Martyn Day put it directly:
“The most valuable thing the ODA owns is not any single format, but its new neutral, self-hosted access to all the standards that matter.”
Like everything ODA builds, this is infrastructure, not an application. A software vendor can ship it inside a product. A firm with its own models can run it in-house and never send a file outside. Neutral either way, and that neutrality is the point.
Early on, reading and analysis work across all three formats. Writing back goes furthest in DWG, and stays lighter in STEP and IFC for now. We would rather be precise about that than oversell it.
If your data cannot leave the building, that is a longer story worth its own read — see the next piece. For the specifics on formats, timing and setup, see the FAQ. Or come talk to us about launch.
