Cad2data Revit IFC DWG DGN

CAD2Data is best understood as the intake layer of a larger open-source construction data stack, not a standalone converter — and that framing is what makes it worth a look. The homepage is explicit that extraction results land in OpenConstructionERP, which carries 55,000+ cost items across 20 interface languages and 18+ modules spanning estimates, takeoff, BOQ editing, tenders, contracts, schedules, site tasks, documents, equipment, safety, validation and reporting. That connected data model is the real argument: a Revit or IFC element gets linked in bulk to a cost item instead of being retyped, and site tasks can hang off the same model. If you have ever rebuilt a quantity takeoff by hand because the model and the estimate lived in different tools, that is the workflow this targets. Strengths. The deployment model is honest and unusual for AEC software: pip install openconstructionerp, docker run openconstructionerp, or a curl one-liner, self-hosted and offline-capable, so model files never leave your machines. The licensing is AGPL-3.0 across the flagship platform, which means no per-seat converter tax and no per-model fee. Underneath sits FastAPI, PostgreSQL and "clean Python you can read," so a plugin is an afternoon rather than a roadmap request. AI agents are treated as infrastructure rather than marketing: the vendor publishes an open schema and documented API, ships DDC Skills as a skill library for data extraction through cost estimation, and names Claude and Cursor specifically as ways to wire an agent in. OpenConstructionEstimate adds the resource, time and cost calculation logic so an agent has something deterministic to call. Weaknesses. The seed data flags that the pipeline needs a local Windows machine for the CAD-side work, has no GUI — operations are command-line — and can be slow on large Revit models because of COM automation overhead. There is also no built-in scheduler, so recurring pipelines need cron or a task scheduler bolted on. Those constraints are consistent with what the scrape shows: everything is framed around files, scripts and modules, never around a browser canvas. Beyond the software, the vendor sells human help — an AI Agent Deployment session, custom workflow development, and workshops — which tells you the realistic adoption path involves either internal Python capability or paid hand-holding. Where it fits. Estimation teams that re-run the same extraction weekly and want it owned rather than rented; BIM managers who need bulk model-to-table conversion with data staying on-prem; automation engineers building construction agents against an open schema. Teams in 87+ countries are cited as users, though that figure is vendor-supplied. Where it doesn't. Architects who want to orbit a 3D model. Anyone whose organization is SaaS-first and has no appetite for Docker. Groups needing real-time multi-user collaboration inside the model. And buyers who want a support contract with a turnkey onboarding — the