NaCl (reading as Natrium Chloride) is an open-source full-SDLC framework, and its core innovation is how it solves the hardest problem in agentic development: storing and retrieving project knowledge. Instead of markdown specs that agents must re-read every session, all project knowledge — requirements, architecture decisions, entities, rules, and their relationships — lives in a Neo4j knowledge graph as first-class queryable objects. The measured effect: planning a single use case takes ~550 tokens (one targeted Cypher query) instead of ~150,000 tokens (reading a ~70-file markdown spec, because the agent doesn't know in advance where the relevant facts live) — a 99.6% reduction in context per use case, repeated in every planning session. Around the graph we've built the full tooling for agentic development and testing: 57 skills covering BA → SA → TDD development → review → QA → release, quality gates before production, and an autonomous goal orchestrator (/nacl-goal). The framework answers "how is this built and why" for any part of the system, regardless of project scale or iteration count. Benchmarked on a real production project (EV charging station management): a classical team was estimated at 5,831 person-hours; running fully on NaCl the same scope takes 1,480 person-hours — 4x fewer person-hours and 60% lower cost.