AI-Powered Construction Estimation - Transform project plans into accurate, risk-adjusted cost estimates using 7 specialized deep agents.
TrueCost is an AI-powered construction estimation system built as a brownfield pivot from the existing CollabCanvas collaborative canvas application. The system employs 7 specialized deep agents that collaborate through structured workflows to transform CAD plans and project descriptions into comprehensive, professionally-credible estimates.
Target Users: General contractors and subcontractors who need fast, accurate estimates to win bids and manage project costs.
- 7-Agent Deep Pipeline - Specialized agents for clarification, CAD analysis, location intelligence, scope generation, cost estimation, risk analysis, and final synthesis
- CAD Plan Analysis - Upload PDF, DWG, or image files for automatic measurement extraction
- Voice + Text Input - Describe projects via natural language (typed or spoken)
- Risk-Adjusted Estimates - Monte Carlo simulation provides P50/P80/P90 confidence intervals
- Professional PDF Output - Generate client-ready estimate reports
- Multi-Retailer Price Comparison - Find optimal material prices across Home Depot, Lowe's, and more
Phase: Implementation Planning (MVP)
The project is transitioning from CollabCanvas (collaborative canvas) to TrueCost (AI estimation engine). See /docs for detailed planning artifacts:
- PRD: docs/prd.md - Product requirements (78 FRs)
- Architecture: docs/architecture.md - Technical design decisions
- Epics: docs/epics.md - Implementation breakdown (5 epics, 24 stories)
- Sprint Status: docs/sprint-artifacts/sprint-status.yaml - Current progress
Frontend (Existing - to be extended):
- React 19 + TypeScript + Vite
- Konva.js for canvas rendering
- Zustand for state management
- Tailwind CSS + shadcn/ui
- Firebase SDK
Backend (New - Python Cloud Functions):
- Python Deep Agents (deepagents 0.2)
- OpenAI GPT-4.1 (configurable via env var)
- Firebase Cloud Functions (2nd gen)
- Firestore + Firebase Storage
- WeasyPrint for PDF generation
Clarification → CAD Analysis → Location → Scope → Cost → Risk → Final
Agent Agent Agent Agent Agent Agent Agent
Each agent:
- Receives structured input from previous agents
- Performs specialized analysis
- Writes output to Firestore (real-time UI updates)
- Hands off to next agent in sequence
- Node.js 20+
- Python 3.11+
- Firebase CLI (
npm install -g firebase-tools)
# Frontend dependencies
cd collabcanvas
npm install
# Python Cloud Functions
cd ../functions
python -m venv venv
source venv/bin/activate
pip install -r requirements.txt
# Environment setup
cp .env.example .env.local
# Edit .env.local with your API keys
# Start emulators
firebase emulators:start
# Start frontend dev server (new terminal)
npm run devLLM_MODEL=gpt-4.1
OPENAI_API_KEY=sk-...
LANGSMITH_API_KEY=ls-... # Optional - for agent tracingtruecost/
├── collabcanvas/ # React Frontend (existing, to be extended)
│ └── src/
│ ├── components/
│ │ ├── estimate/ # NEW: TrueCost UI components
│ │ └── ... # Existing canvas components
│ ├── hooks/
│ ├── services/
│ └── stores/
│
├── functions/ # Python Cloud Functions (NEW)
│ ├── agents/ # 7 Deep Agents
│ ├── services/ # CAD parsing, Monte Carlo, PDF generation
│ └── templates/ # PDF report templates
│
├── docs/ # Documentation
│ ├── prd.md # Product requirements
│ ├── architecture.md # Technical architecture
│ ├── epics.md # Epic/story breakdown
│ └── sprint_artifacts/ # Sprint tracking
│
└── bmad/ # BMAD workflow system
| Document | Purpose |
|---|---|
| PRD | Product requirements, success criteria, functional specs |
| Architecture | Technical decisions, data models, API contracts |
| Epics | Implementation breakdown, FR coverage, verification checklists |
| Product Brief | Problem statement, vision, competitive analysis |
The TrueCost pivot leverages existing CollabCanvas infrastructure:
- Firebase Auth (Google OAuth)
- Firestore data persistence patterns
- Cloud Functions for AI integration
- Firebase Storage for file uploads
- React component library
- Zustand state management
- Project management CRUD
- BOM infrastructure
See docs/archive/collabcanvas-stories/ for historical CollabCanvas implementation details.
This project uses the BMAD Method for planning and implementation:
- Use
/bmad:bmm:workflows:create-storyto draft stories from epics - Use
/bmad:bmm:workflows:dev-storyto implement stories - Use
/bmad:bmm:workflows:sprint-planningto track progress
This project is licensed under the MIT License.
TrueCost - Accurate estimates in minutes, not hours.