LAAP
An open-source cognitive architecture, written mostly in Python with an optional Rust core, that models perception, memory, and causal reasoning without routing every step through a large language model, meant to run alongside an LLM agent such as Hermes.
LAAP (Living Agent Application Protocol) is a cognitive architecture for autonomous agents built from independent modules rather than a single LLM call: a PSI Core for simulated physiological needs, a quantum-reasoning vector engine, an episodic memory store with a knowledge base, a rules engine, and separate causal-reasoning, analogical-reasoning, and world-simulation engines. Its authors describe the approach as "Zero-LLM": most cognitive functions are handled by this architecture rather than by language generation, with an LLM treated as one component rather than the whole system.
This repository is the first open release: an AGI cognitive engine, an integration layer for the Hermes agent, and a pure-Python fallback for the PSI Core that runs without the Rust binary. A native Rust PSI core, front-end clients, and further harness components are not included here.
The project follows an open-core model: this repository is Apache 2.0, while enterprise multi-agent orchestration, cloud cognitive sync, federated learning, and additional domain modules require a commercial license.