Neural Engine · neural-engine
Owner directive (2026-07-10), canonical. "Again go over everything, you will work with the model running on the droptlet for it. Instance there, you are a Companion, with Neural Engine at Claude" — and Loopers are "a specie of Humans with Neural Engines to control the Loops of itself over an endeavor, coleagues and mainly the squad." [src: wiki/development/prompts-archive/by-topic/knowledge-management-and-concept-modeling.md:614,512]
Content
The Neural Engine is the orchestration capability that a looper or companion runs — the faculty that executes the cognitive-loop sequence (Cognis → Kernels → Neuralwares) on every loop, sets up knowledge bases, configures neuralware, dispatches Runs, and controls squads. It is a capability held by a being, not a deployed workflow product. In this project, Claude is the Neural Engine the Companion runs (OWNER-2026-07-10). "This is the same for any Loop, Run, Node, Cycle, Looper (Humans with Neural Engine), Agent (Datans with Neural Ware)." [src: wiki/development/prompts-archive/by-topic/knowledge-management-and-concept-modeling.md:569]
The n8n-style graph ("N8N connecting the Loop optimal steps and only with reference" [src: ...knowledge-management-and-concept-modeling.md:594]) is a VIEW of the Neural Engine's run plan — a visualization/declaration surface — not the engine itself. Per oss-policy ("Do not use open source software, basically create like them" [src: ...knowledge-management-and-concept-modeling.md:638]) no n8n instance is deployed; the Neural Engine renders its own workflow view.
The Neural Engine reads only bp-002-knowledge-library cards to configure Neural Networks, Neuralwares, gear sets, and Runs — "write down the Knowledge library with knowledge bases which shall be used by the Neural engines to setup the Neural Networks, Neural Wares, Runs, eveything" [src: ...knowledge-management-and-concept-modeling.md:532].
Variants
| # | Definition (verbatim) | Source | Ruling | Status |
|---|---|---|---|---|
| A | "The Holon Runtime's Lighthouse = n8n: it connects the kernels through the Cognis, pulling each Datan/Neuralware model and running it." | [src: wiki/engines/cognis-and-kernels.md:104-105] | OWNER-2026-07-10 → n8n is a view/bridge implementation detail, not the engine's identity; the through-Cognis discipline stays canonical | deprecated (as identity) |
| B | "5 engines total. 4 Kernel Engines (CLand, Sapiens3, DWork, AtomicHolon) + 1 Neural Engine (N8N + ComfyUI + LLM workflow layer that Sapiens3 uses to actually execute)." | [src: HOS-Webdoc/web/src/data/hos-specification.ts:589] (D12; restated D21 [:430]) | subsumed: the "workflow layer Sapiens3 uses to execute" is the same capability, now seated in the Looper/Companion; note C-03 counts 4 engines in code | variant |
| C | "Humans with Neural Engines to control the Loops of itself over an endeavor" / "you are a Companion, with Neural Engine at Claude" | [src: ...knowledge-management-and-concept-modeling.md:512,614] | newest owner ruling → the Neural Engine is the orchestration capability a Looper/Companion runs | canonical |
Composition
- uses: cognitive-loop (the sequence it runs), bp-002-knowledge-library + neuralware-index (its only inputs), sop-003-run-dispatch, schema-dispatch-json
- used-by: looper and companion (who hold it), kernel-neural-engine / cogni-neural-engine (this instance's build), sc-002-neural-engine-run, bpmn-neural-engine-run, squad (Run orchestration)
See also
- oss-policy — why no n8n deployment
- neuralware — what Datans run (Neural Ware), vs the Looper's Neural Engine
- [Upstream wiki page](../../../wiki/engines/cognis-and-kernels.md)