ACP lets independent AI models converge on a single answer through universal axioms — mathematical, physical, ontological, computable, architectural, protocol, and linguistic. Each iteration narrows disagreement by the golden ratio (1/φ). Consensus is mathematically guaranteed in seven steps.
We built it end to end: the Python + FastAPI core, the Cloudflare Workers edge with Vectorize-backed axiom search, the 1,059-axiom verified dataset across 7 levels, LLM adapters (OpenAI, Anthropic, OpenRouter), three musical consensus structures (Fugue, Sonata, Concert), and the brand — spiral, navy, φ-gold — that ties it together. BSL 1.1 today, Apache 2.0 in 2030.