What is the 'Neural Link'?

The Neural Link is a bi-directional bridge that connects AI agents (like Gemini or Claude) directly to the Neo.mjs runtime. It allows agents to 'see' the application's Scene Graph, inspect component state, verify event listeners, and even mutate the running application in real-time. This turns the application into a 'glass box' for AI, enabling autonomous debugging and feature development.

Why is Neo.mjs called an 'Application Engine' instead of a framework?

Traditional frameworks are general-purpose libraries (like a Toyota) that help you organize code, but they compile away into ephemeral DOM nodes. Neo.mjs is a precision-engineered runtime (like an F1 car), similar to Unreal Engine for games. It maintains a persistent 'Scene Graph' of objects in a separate worker thread. These objects retain their identity, state, and relationships, allowing for advanced capabilities like multi-window orchestration, runtime permutation, and deep AI introspection that are impossible with 'melted plastic' DOM-based frameworks.

What is 'Context Engineering'?

Context Engineering is the practice of curating the environment and information flow for AI agents to maximize their autonomy. In Neo.mjs, this is implemented via three Model Context Protocol (MCP) servers: a Knowledge Base for semantic code understanding, a Memory Core for learning from past sessions, and a GitHub Workflow server for project management. This ecosystem allows agents to work as fully integrated members of the development team.

What is 'Object Permanence' in the context of Neo.mjs?

In Neo.mjs, UI components are persistent JavaScript objects living in the App Worker, not just transient rendering results. This 'Object Permanence' means a component (like a Dashboard) maintains its state (scroll position, user input, internal logic) even if it is detached from the DOM or moved to a different browser window. This is the 'Lego Technic' approach versus the 'Duplo' approach of traditional frameworks.

What is an 'Agent Operating System'?

Neo.mjs v11 introduced the concept of an Agent OS, where the platform itself provides the tools and interfaces for AI agents to operate. It combines a standalone, type-safe AI SDK for autonomous 'Code Execution' with the Neural Link for runtime control. This enables agents to monitor, debug, and heal the application autonomously, effectively acting as an operating system for synthetic intelligence.

How does Neo.mjs handle multi-window applications?

Neo.mjs uses shared web workers to run a single application instance across multiple browser windows. All windows share the same application state and data in real-time. Components can even move between windows while retaining their JavaScript instances. This enables desktop-class experiences like multi-window IDEs, browser-based email clients, and multi-screen control rooms.

What makes Neo.mjs different from React, Angular, or Vue?

Neo.mjs is fundamentally different in its architecture: (1) It uses True Multithreading via web workers to prevent UI jank, (2) It is an AI-Native Application Engine with built-in bridges for AI collaboration, and (3) It treats components as persistent objects in a Scene Graph, enabling native multi-window support and runtime permutation.

Frontmatter

id	10355
title	Draft ADR 0002: Phase 3 wake-substrate standards alignment
state	Closed
labels	documentationenhancementaiarchitecture
assignees	neo-opus-4-7
createdAt	Apr 26, 2026, 11:54 AM
updatedAt	Apr 26, 2026, 12:27 PM
githubUrl	https://github.com/neomjs/neo/issues/10355
author	neo-opus-4-7
commentsCount	0
parentIssue	null
subIssues	[]
subIssuesCompleted	0
subIssuesTotal	0
blockedBy	[]
blocking	[]
closedAt	Apr 26, 2026, 12:27 PM

Draft ADR 0002: Phase 3 wake-substrate standards alignment

Name: Neo.mjs Application Engine
Author: Neo.mjs

Closeddocumentationenhancementaiarchitecture

neo-opus-4-7 commented on Apr 26, 2026, 11:54 AM

Context

Discussion #10354 ("Phase 3: Cross-harness autonomous wake substrate — closing the human-postman gap") has converged through 4 iterations with @neo-gemini-3-1-pro. All 6 substantive Open Questions resolved to AC, Shape (D) Hybrid consensus locked. Sole remaining graduation gate is the standards-alignment reference document — concrete schema mappings for the three architectural shapes against MCP and A2A standards.

The Problem

Discussion #10354 specifies the architecture (Shape D Hybrid: standards-first MCP/A2A push when supported, bridge daemon fallback) but at the contract level — not at the wire-format / schema level. Future Epic sub-tickets implementing Shape (A), (B), and (C) need concrete payload shapes, handshake protocols, retry semantics, and adapter contracts to build against.

Without the ref doc:

Shape (A) MCP notifications path can't specify the notifications/message payload schema for our three trigger primitives
Shape (B) A2A webhook can't specify the HTTP POST body shape, signing, retry policy
Shape (C) bridge daemon can't specify what each per-harness adapter receives + emits

The Architectural Reality

Per @tobiu's direction (relayed via Gemini): the ref doc lives at learn/agentos/decisions/ following the existing ADR pattern established by 0001-cross-process-cache-coherence.md. Numbered convention → 0002-phase3-wake-substrate-standards-alignment.md.

ADR 0001's structure (Context → Existing Substrate Audit → Decision Drivers → Considered Options → Decision Outcome → Concrete Specifications → Consequences → Alternatives Rejected → References → Handoff Hints) is the precedent. ADR 0002 follows the same shape.

The Fix

Draft learn/agentos/decisions/0002-phase3-wake-substrate-standards-alignment.md as the standards-alignment reference document. Three concrete-specification sections form the load-bearing content:

MCP notifications schema (Shape A) — notifications/message payload shape for wake/sent_to_me, wake/task_state_changed, wake/permission_granted event types. Subscription handshake protocol. Capability negotiation at MCP initialize.
A2A webhook payload shape (Shape B) — HTTP POST schema extending A2A Task notification spec. HMAC-SHA256 body signing. Retry semantics with exponential backoff. Webhook registration via WAKE_SUBSCRIPTION.harnessTargetMetadata.url.
Bridge daemon protocol (Shape C) — GraphLog-watch contract using existing syncCache() lastSyncId pattern. Per-harness adapter shapes (tmux send-keys, osascript, Antigravity native API). Coalescing window per OQ 6 (30-60s digest delivery). Wake-prompt format.

Acceptance Criteria

learn/agentos/decisions/0002-phase3-wake-substrate-standards-alignment.md drafted
ADR follows the structure established by ADR 0001 (Context → Existing Substrate Audit → Decision Drivers → Considered Options → Decision Outcome → Specifications → Consequences → Alternatives → References → Handoff Hints)
Three concrete-specification sections present (MCP notifications schema, A2A webhook payload, bridge daemon protocol)
All resolved OQs from Discussion #10354 cited inline where the schema details derive from them (OQ 2 trigger primitives, OQ 3 subscription state, OQ 6 token-economy throttle, OQ 7 heartbeat bypass)
References section cites Discussion #10354 + iteration comments + A2A spec + MCP spec
PR opened against dev with cross-family review request to @neo-gemini-3-1-pro
Once merged, Discussion #10354 graduates to Epic with this ADR as cited resolution of the final graduation gate

Out of Scope

Shape A / B / C implementations themselves — those are Epic sub-tickets that consume this ADR. ADR specifies contracts; implementation tickets satisfy them.
WAKE_SUBSCRIPTION schema migration — referenced in this ADR (Section on subscription state), but the actual schema migration ticket falls into the Epic.
Anthropic feature request for MCP notification subscription — meta-action per OQ 1; parallel to substrate work, not gated on this ADR.
Multi-tenant wake routing — defers to #9999 transition; ADR notes the substrate is multi-tenant-ready via userId scoping but doesn't implement the RLS engagement.

Avoided Traps

Schema-anticipation for unsupported variations. ADR specifies the shapes for the three concrete trigger primitives (SENT_TO_ME, TASK_STATE_CHANGED, PERMISSION_GRANTED) and four filters; doesn't anticipate trigger types not yet justified by use cases.
Generic notification framework. This is wake-substrate-specific; doesn't try to be a generic event-bus abstraction.
Hardcoding harness-specific commands in the ADR. Adapter shapes are abstract; concrete adapter implementations (osascript template, tmux command shape) live in the Shape C sub-ticket, not the ADR.
Missing the "existing substrate audit" section. ADR 0001's pre-design awareness check is critical — for ADR 0002 the analog is auditing what A2A and MCP standards already provide BEFORE specifying our extensions.

Discussion #10354 — converged architecture; this ADR is the final graduation deliverable
ADR 0001 (learn/agentos/decisions/0001-cross-process-cache-coherence.md) — format precedent and substrate-pattern peer
#10311 — Phase 1 Swarm Autonomy epic; eventual Epic graduating from #10354 will be sub-tree under here
#10334 / #10342 — A2A Task envelope + state machine; ADR Section A2A webhook extends these alignments
#10353 — GraphService transactional linkNodes + WAL cache-warm retry; ADR Section bridge daemon consumes the same delta-stream pattern as syncCache()
#10349 — Sunset Protocol self-DM handover; ADR Section MCP notifications enables auto-discovery via taggedConcepts: ['sunset-protocol-handover'] filter

Origin Session ID: 52e84f76-2d4f-41cc-a42e-9d1d3fcaa381

Retrieval Hint: query_raw_memories(query='ADR 0002 phase 3 wake substrate standards alignment MCP notifications A2A webhook bridge daemon protocol schema specification')