Pitchforge: A Multi-Agent LangGraph Framework with Retrieval-Augmented Generation for Comprehensive Startup Viability Analysis | IJCT Volume 13 – Issue 3 | IJCT-V13I3P10

The evaluation and planning of new business ven- tures remain a complex, high-risk endeavor traditionally depen- dent on human expertise. The rapid expansion of global startup ecosystems has intensified the need for intelligent decision- support systems capable of performing comprehensive startup viability analysis autonomously. Recent advances in Large Lan- guage Models (LLMs) have catalyzed a paradigm shift, introduc- ing data-driven methodologies for market research, cost estima- tion, and strategic planning. However, existing approaches largely remain limited to single-domain tools, leaving a critical gap in systems capable of autonomously generating comprehensive, actionable business analyses from a nascent idea. This paper introduces IdeaArchitect AI, a multi-agent system built using the LangGraph framework that orchestrates special- ized LLM agents to address this gap. The framework integrates seven domain experts covering market research, cost estimation, legal compliance, technology architecture, monetization strat- egy, government schemes, and strategic planning. A Retrieval- Augmented Generation (RAG) pipeline based on ChromaDB enables domain-specific knowledge grounding using sentence- transformer embeddings. The system adopts an Orchestrator– Specialist–Critic architecture where an orchestrator dynamically selects agents, specialists perform domain analysis, and a critic performs adversarial validation before final refinement. Experimental evaluation demonstrates that the system achieves expert ratings comparable to professional consulting outputs while completing full analyses in under four minutes on com- modity hardware. Results confirm that coordinated multi-agent LLM systems represent the next logical frontier in agentic AI, capable of significantly enhancing decision-support capabilities for early-stage entrepreneurs.