Umbra-Noesis · Protocol Layer

AACP

The language your AI
and your system
actually share.

AACP is the protocol layer that turns intent into structure. It defines how humans, local AI systems, tools, and memory exchange information in a stable, deterministic, auditable way — across every node in the Umbra-Noesis stack.

Deterministic Governed Licensable Auditable

AACP Protocol Stack

AACP Core Foundation

Core message schema — fields, structure, and contract for every human–AI exchange.

Context & State

Tracks environment, task, and session — responses stay grounded.

UN-SLP Symbol Layer

Lightweight symbol tags for mode, tone, and intent.

Local Ruleset Bridge

Connects messages to node-specific safety and policy.

Tool & Runner Binding

Encodes which tools or workflows execute per message.

MMC — Memory & Continuity Long-Term

Distributed memory and continuity channels that ride on top of AACP routes across nodes and time.

AACP Core — active in EVE OS
UN-SLP Symbol Layer — active
Reference implementation — licensable

The Problem

Prompting is guesswork. AACP is structure.

A prompt is an unstructured natural language request to a system that doesn't know what you mean — only what you said. The system guesses at intent, guesses at context, guesses at what tool to use. AACP eliminates that guesswork by defining the contract both sides of the conversation actually follow.

Every AACP message carries structured fields: who is sending, what context is active, what the intent is, what rules apply, and what the expected response form is. Nothing is inferred. Nothing is ambiguous. The communication is deterministic by design.

Structured message contracts — not free-form natural language

Context binding — session, task, and environment travel with every message

Policy enforcement at the protocol level — not bolted on afterward

Auditable — every exchange follows a known schema, traceable after the fact

Layer Breakdown

AACP Core

The base schema that defines every message field — sender, intent, context reference, tool binding, and response contract. Everything else extends from this layer.

Context & State Layer

Tracks the active environment, current task, and session state so every response is grounded in what is actually happening — not just what was last said.

UN-SLP Symbol Layer

Lightweight symbol tags for communication mode, tone, and intent that sit above AACP Core. Enables fast, low-overhead modifiers without restructuring the base message.

Tool & Runner Binding

Encodes exactly which tools, runners, or workflows should execute for a given message. No ambiguity about what should happen — the protocol carries the instruction.

Licensing

The most portable layer in the stack

AACP is designed to operate independently of the full Umbra-Noesis environment. It can enhance existing AI deployments without requiring EVE OS, EVAA, or PCe hardware. If you have a human–AI workflow that needs structure and governance, AACP is the layer to license.

Protocol Specification

Full message schema, field definitions, and implementation reference.

Runner Shells

AACP-native orchestration layers for local and cloud model deployment.

Memory & UI Components

Injectors, structured tools, and interfaces that speak AACP natively.

Partnership & Licensing

Question everything.

Especially the way AI systems are deployed, owned, and controlled. Umbra-Noesis exists to give builders and operators a different option: local-first, protocol-governed, and unapologetically human-centric.

For licensing or partnership inquiries:

umbra-noesis@pm.me

The shift

Quit prompting.
Start communicating.

Prompting is guesswork with a chat interface. AACP is a shared language — structured, bounded, and governed from both ends. The difference is between hoping the AI understood you and knowing it did.