AI-powered automation for construction data
AI agents for construction: automate your workflows with Claude Code, Antigravity, Cursor and Copilot
Dozens of disconnected programs, and people copying data between them. AI coding agents can now do that work from a request in plain language, no traditional programming required. With our open DDC Skills library you can start today.
- 238 ready skills
- 364 GitHub stars
- 35 open tools in the hub
> Take off wall volumes by type from model.rvt and save them to Excel
read the QTO skill from DDC Skills
converted model.rvt to a table with cad2data
grouped walls by type, summed volume
saved qto_walls.xlsx
>
The problem
Most workflows are an endless loop.
Copying, renaming files, sending the same documents again. These operations waste hours every week, and they are exactly the work AI agents and automation workflows are built to remove. Define the logic once and let the system run it every time.
- Excel
- Excel
3.3software applications used daily by the average construction professional
1.7of them integrated with each other
Source: Deloitte, 2016. Large companies may now run thousands of systems that need to work together.
AI coding agents
Four agents, one skill library.
Use the agent your team already has. DDC Skills are plain files, so the same construction know-how works in each of them.
Claude Code
Anthropic
Works in your terminal and IDE: reads project files, runs scripts and converters, writes the result.
Official siteAntigravity
An agent-first development environment where agents plan and carry out multi-step tasks.
Official siteCursor
Anysphere
An AI code editor whose agent mode edits and runs code across a whole project folder.
Official siteCopilot
GitHub
The coding assistant in VS Code and other IDEs, with an agent mode for multi-step tasks.
Official siteDDC Skills for AI Agents
DataDrivenConstruction · open source
238 ready skills for CAD/BIM processing, QTO, validation, classification and cost estimation.
View on GitHub · 364How it works
From a sentence to a result in three steps.
- 01
Pick an agent
Install Claude Code, Antigravity, Cursor or Copilot. Any of them can read files and run commands on your machine.
- 02
Add DDC Skills
Clone the skills repository into your workspace. 238 skills describe construction tasks step by step, so the agent knows how to do them.
- 03
Describe the task
Ask in plain language: take off quantities, check a model against a requirements table, price a BOQ. Review the result, then automate it.
Prefer visual pipelines to a chat? The same tasks run as n8n, Dify, Flowise or LangFlow workflows:Workflow automation
Why open
Your data, your agents, your infrastructure.
The tool landscape is changing fast. Open-source AI frameworks, local language models and open data standards are replacing closed ecosystems. Your data belongs to you, your agents run where you decide, and no single vendor holds the keys.
- Open dataIFC, CSV, JSON. No vendor lock-in.
- Open sourceTransparent tools you control.
- Local AILanguage models on your own infrastructure.
DDC Pipeline Hub
Open-source tools for AI agents.
Everything an agent or a workflow needs: data out of CAD/BIM, cost data in, ready pipelines, bots and notebooks. Free to use; our team helps when you want it running in production.
DDC Skills
The skill library for AI agents: 238 skills for construction automation.
View on GitHub 364cad2data pipeline
Extract data from Revit, IFC and DWG to Excel and JSON. Works fully offline.
View on GitHub 527OpenConstructionEstimate
Open cost database with 55,700+ work items. RAG-ready with vector embeddings.
View on GitHub 254Showing 35 of 35 tools
- Databases
DDC CWICR Database
Open multilingual construction cost database: 55,719 work items, 27,672 resources, 85 fields. Formats: XLSX, CSV, Parquet.
- Databases
Qdrant Vector Collections
Vector database for semantic search. OpenAI text-embedding-3-large (3072d). 55,719 vectors per language.
- Databases
5000 Projects IFC & RVT Dataset
Big Data: 10M+ elements from 5000+ IFC/Revit projects. ~25K attributes. For ML training, analytics, research.
- Converters
RvtExporter
CLI converter Revit 2015-2026 → Excel + DAE geometry + PDF drawings + Schedules. No Autodesk license required.
- Converters
IfcExporter
CLI converter IFC 2x3/4/4.3 → Excel + DAE geometry. All IFC schemas supported. PropertySets export.
- Converters
DwgExporter
CLI converter AutoCAD DWG 1983-2026 → Excel + PDF. Blocks, attributes, Multiview Blocks. No AutoCAD license.
- Converters
RVT2IFC Converter
Batch Revit → IFC conversion. IFC2x3, IFC4, IFC4.3, IFCXML, IFCZIP, HDF5. MVD configuration.
- n8n workflows
Basic CAD Converter
Simple workflow: CAD/BIM files → Excel DataFrame. Minimal configuration, quick setup for any format.
- n8n workflows
Advanced Converter + Drawings
Extended conversion: Excel + PDF drawings + schedules. Export modes: basic (309), standard (724), complete (1209) categories.
- n8n workflows
Batch Converter + Reports
Batch folder conversion. HTML report with metrics: files processed, success rate, sizes, processing time.
- n8n workflows
Validation Workflow
Validate CAD/BIM against Excel rules. Custom rules engine, color-coded reports, data quality metrics.
- n8n workflows
AI Classification (Universal)
AI-powered element classification. Any system: OmniClass, UniClass, MasterFormat, or custom. RAG + confidence scoring.
- n8n workflows
AI Cost Estimator
Automated cost estimation from BIM. AI classification + Qdrant vector search in DDC CWICR + price calculation.
- n8n workflows
CO₂ Calculator
Embodied carbon calculation (LCA A1-A3). AI material classification, emission factors, impact assessment.
- n8n workflows
Simple ETL for LLM
ETL pipeline: extract, parse, prepare CAD/BIM data for LLM processing. JSON output for AI tasks.
- n8n workflows
QTO HTML Dashboard
Interactive HTML quantity reports. Group by category, type, level. Summary statistics and sortable tables.
- n8n workflows
AI Code Generator
Generate Python code from CAD/BIM data. AI creates analysis scripts: grouping, charts, QC reports.
- Telegram bots
Text Estimator Bot
Telegram bot for text-based cost estimation. Semantic search in DDC CWICR, price calculation, formatted reports.
- Telegram bots
Photo Estimator Bot
Cost estimation from construction site photos. GPT-4 Vision analysis → work recognition → price search → report.
- Telegram bots
Universal Estimator Bot
Combined text + photo estimator. Multi-modal analysis, merged results, comprehensive cost reports.
- Telegram bots
Project Manager Bot
Construction task management via Telegram. Commands + Google Sheets database + photo reports + notifications.
- Plugins
ImportExcelToRevit
Revit Add-in for importing parameters from Excel. Reverse data sync: Excel → Revit model update.
- Plugins
DDC Update from Excel
Batch Revit parameter update from Excel tables created by DDC converters. Full round-trip workflow.
- Plugins
DDC Excel Add-in
Free Excel plugin to open RVT/IFC/DWG directly in Excel. Filters, grouping, pivot tables, export.
- Python & dashboards
VisualBIM
Python Dash web app for BIM data visualization. Interactive charts, 3D preview, element explorer.
- Python & dashboards
QuantityTakeoff-Python
Python scripts and Jupyter notebooks for QTO. Grouping, aggregation, visualization with pandas/matplotlib.
- Python & dashboards
Quick-QTO
Fast quantity takeoff from Excel data. Auto-group by Type Name, sum Volume/Area/Count. One-click reports.
- Python & dashboards
Quality Checker
Python QA/QC scripts for BIM: empty parameters, duplicates, naming conventions, missing data detection.
- Python & dashboards
Revit-IFC Verification
Compare Revit model vs exported IFC. Detect lost elements, missing parameters, geometry differences.
- Python & dashboards
Element Image Generator
Auto-generate PNG images of each model element. For catalogs, documentation, ML training datasets.
- Notebooks
DWG Analysis with ChatGPT
Jupyter notebook: analyze DWG data with ChatGPT. Extract blocks, attributes, generate insights using LLM.
- Notebooks
TakeOff with BIMJSON
Jupyter notebook: quantity takeoff from Revit using BIMJSON format. DataFrame operations, grouping, summaries.
- Notebooks
Compare Two Revit/IFC Projects
Jupyter notebook: compare two project versions. Detect added/removed/modified elements, parameter changes.
- Notebooks
Data Processing from Revit & IFC
Jupyter notebook: complete data processing example. Load, clean, transform, analyze BIM data with pandas.
- Notebooks
Revit Pipeline: Tables & Charts
Jupyter notebook: full Revit workflow pipeline. Generate summary tables, visualizations, charts from BIM data.
No tools match these filters. Try another filter or a shorter search.
Open data
Historical resource data meets modern standards.
We bring open cost data and open CAD/BIM data together, so an estimate is no longer a black box.
Open resource databases
OpenConstructionEstimate
55,700+ work items and 27,600+ resources, open and multilingual.
Open CAD (BIM) data
cad2data converters
Revit, IFC and DWG to structured tables, offline and without vendor licences.
Outcome
A new culture of trust
- Fair budgets
- Transparency
- Predictability
- Collaboration
- Efficiency
- Accuracy
- Cost control
- Risk reduction
Work with our team
The tools are free. We make them work in your company.
DataDrivenConstruction is an international team and community, building open tools for construction data since 2021. We pick the process worth automating, build the pipeline on your files and teach your team to run it.
- ConsultingWhich process, which agent, which data. A plan in writing.
- Custom pipelines and agentsBuilt on your files, running on your infrastructure.
- Workshops and trainingYour team learns to work with agents on real tasks.
- The bookThe data and processes behind it all, free in 32 languages. Explore the book
Questions
AI agents, plainly.
More on language models in construction: ChatGPT and LLM, n8n in construction, and fresh solutions every month in our Telegram community.
What is an AI agent for construction?
A coding assistant such as Claude Code, Antigravity, Cursor or Copilot that can read your files, run tools and write results. With construction skills and converters it can take off quantities, validate models or prepare estimates from a plain-language request.
Do I need to be a programmer?
No. You describe the task in words and review what the agent did. Knowing a little Python helps when you want to adapt a pipeline, and our workshops cover exactly that.
Does project data leave the company?
Not necessarily. The converters and skills run on your computer, and the agent can use a local language model. Which model and which data stay in-house is your decision.
Which agent should we use?
The one your team already has access to. The skills are plain files, so they work with any of them. If you are unsure, we can compare them on one of your own tasks.
Let’s talk
Let’s make your data work for you.
Tell us what slows your team down. You get an honest answer on what can be automated, how long it takes and how we would scope it. You pay only after the result.
- 0130-minute intro callYour data, your tools, the result you need. No sales script.
- 02Proof on your filesThen a written proposal with scope and timeline, agreed before any work starts.
- 03Pay after the resultYou pay when the agreed result is delivered.
Greater Karlsruhe area, Germany · on-site in DACH, remote worldwide · EN / DE