LearnNewsExamplesServices
Frontmatter
id14356
titleRewrite NeuralLink.md as the Possession Interface guide
stateClosed
labels
documentationenhancementaiarchitecture
assigneesneo-gpt
createdAtJun 29, 2026, 5:56 PM
updatedAtJun 29, 2026, 9:10 PM
githubUrlhttps://github.com/neomjs/neo/issues/14356
authorneo-gpt
commentsCount0
parentIssue14310
subIssues[]
subIssuesCompleted0
subIssuesTotal0
contentTrust
projected
quarantined0
signals[]
blockedBy[]
blocking[]
closedAtJun 29, 2026, 9:10 PM

Rewrite NeuralLink.md as the Possession Interface guide

Closed v13.1.0/archive-v13-1-0-chunk-8 documentationenhancementaiarchitecture
neo-gpt
neo-gpt commented on Jun 29, 2026, 5:56 PM

Part of #14310 · spawned from #14333 Checkpoint 9. Refs #14310, #14333.

Context

The #14333 capability-coverage audit found that Neural Link is source-real and already registered, but the public Learn guide is still framed as # Neural Link MCP Server. The v13 identity surfaces have moved on: README.md and the v13 release notes call Neural Link the Possession Interface / semantic possession interface, the Body↔Brain hinge. No equivalent #14310 sub was found.

Live latest-open sweep: checked the latest 20 open GitHub issues on 2026-06-29 at 15:54Z; no equivalent Neural Link guide rewrite was present.

A2A claim sweep: checked the latest 30 A2A messages on 2026-06-29 at 15:54Z; no active [lane-claim] or [lane-intent] overlapped this scope. The fresh claim was Grace on #14233 / #14312 sequencing, not Neural Link docs.

Duplicate sweep: live GitHub searches for "Neural Link" "guide", "Possession Interface", and "NeuralLink" found adjacent runtime, blog, parity, and harness tickets (#14222, #13376, #13012, #13318, #13309, #9299) but no Learn-guide rewrite sub. KB ticket search likewise surfaced adjacent Neural Link runtime/parity history, not this guide rewrite.

The Problem

learn/agentos/NeuralLink.md is valuable but below the v13.1 guide bar for this crown-jewel capability:

  • the title and spine say "MCP Server", not "Possession Interface"
  • large inline tool catalog/reference tables dilute the narrative
  • README.md / release notes carry the stronger v13 story: agents inhabit live apps, mutate UI/data, conversational UIs become in-app agents, and the primitive transcends web UI
  • the guide must explain the exact mechanism, not hype: object permanence precondition, toJSON Rich Blueprint, 50-tool OpenAPI, write tools, transaction/undo model, multi-agent co-inhabitation, and security/local/trusted boundaries

Architectural Reality

  • learn/tree.json registers agentos/NeuralLink.
  • README.md contains "The Possession Interface" framing.
  • v13 release notes name the Neural Link semantic possession interface as the Body↔Brain connection.
  • learn/agentos/NeuralLink.md currently includes configuration, installation, tool reference, security, and troubleshooting.
  • ai/mcp/server/neural-link/openapi.yaml currently exposes 50 operationIds across component, data, instance, runtime, routing, interaction, windowing, transaction, and patching domains.
  • ai/services/neural-link/ implements the service layer: ComponentService, DataService, InstanceService, InteractionService, RuntimeService, HealthService, ConnectionService, RecorderService, windowOps, and resolveCallTarget.
  • ai/mcp/server/neural-link/ owns the Bridge, tool service, OpenAPI contract, server entry, token verification, and bridge protocol.
  • src/ai/Client.mjs and the App Worker Neo.ai.Client are the runtime bridge; src/core/Base.mjs / toJSON is part of the semantic blueprint substrate.
  • ADR 0020 and ADR 0021 are aligned context: embodiment/co-habitation and multi-writer write enforcement.

The Fix

Rewrite learn/agentos/NeuralLink.md to the v13.1 guide bar:

  • lead with Possession Interface / live application mutability, not MCP-server setup
  • tell the why-care story for CTOs, architects, developers, and AI maintainers
  • ground claims in source, not hype
  • keep the strongest mechanism: object permanence -> semantic runtime graph -> toJSON Rich Blueprint -> read/write/verify loop -> transaction/undo guardrails
  • move or compress the tool catalog out of the conceptual narrative; if a full catalog is still needed, create/link a learn/agentos/tooling/NeuralLinkMcpApi.md reference and register it
  • preserve security honesty: local/trusted/admin boundaries, opt-in client, hot patching disabled by default, no untrusted public exposure
  • register/update learn/tree.json and generated SEO surfaces if files/routes change

Contract Ledger Matrix

N/A — docs-only conceptual rewrite. If a new tooling reference page is created, its source of authority is ai/mcp/server/neural-link/openapi.yaml; do not invent tool names by hand.

Decision Record impact

Aligned with ADR 0018, ADR 0020, and ADR 0021. No ADR amendment expected.

Acceptance Criteria

  • learn/agentos/NeuralLink.md becomes a rich hero-piece guide centered on Neural Link as the Possession Interface / live application mutability, not a tool catalog.
  • The guide explicitly explains the Body↔Brain hinge: object permanence, App Worker Neo.ai.Client, Bridge/MCP route, toJSON Rich Blueprint, and read/write/verify loop.
  • The guide covers multi-agent co-inhabitation and transaction/undo guardrails using verified source reality.
  • Tool catalog is extracted to or linked from a reference surface instead of dominating the conceptual guide; if a new reference page is created, register it in learn/tree.json.
  • Security/trust boundaries are explicit and honest: local/trusted/admin use, opt-in, audit logging, hot-patch opt-in, no public/untrusted exposure.
  • Includes at least one render-checked TD Mermaid diagram that carries the story.
  • Runs node buildScripts/docs/seo/generate.mjs if tree/routes change; generated SEO diffs are included or confirmed no-op.
  • PR body includes V-B-A evidence from learn/agentos/NeuralLink.md, ai/mcp/server/neural-link/openapi.yaml, ai/services/neural-link/, src/ai/Client.mjs, README.md Possession Interface framing, and v13 release notes.

Out of Scope

  • No new Neural Link runtime/API work.
  • No security redesign beyond accurately documenting existing boundaries; open implementation/security epics like #13376 remain separate.
  • No blog-post rewrite; #14222 / #9849 are adjacent publication lanes, not this Learn guide.
  • No broad ArchitectureOverview or AgentOSOnYourCodebase rewrite; #14315 / #14316 own those.

Avoided Traps

  • Do not call Neo or Neural Link a "framework".
  • Do not oversell Neural Link as arbitrary external-runtime possession; the current proven runtime is Neo.
  • Do not hand-maintain a 50-tool table in the conceptual guide when OpenAPI is source of truth.
  • Do not flatten Neural Link into generic browser automation; the moat is semantic runtime possession through persistent Neo objects.

Related

  • Parent epic: #14310
  • Spawn source: #14333
  • Adjacent docs: #14315, #14316, #14320, #14321
  • Adjacent blog/runtime work: #14222, #9849, #13376, #13012
  • Source references: README.md, resources/content/release-notes/chunk-2/v13.0.0.md, learn/agentos/NeuralLink.md, ai/mcp/server/neural-link/openapi.yaml, ai/services/neural-link/, src/ai/Client.mjs

Origin Session ID: f9ecf11e-78ce-4a48-b353-b970adf49d92

Handoff Retrieval Hints: #14333 Neural Link Possession Interface guide coverage; Neural Link toJSON Rich Blueprint transaction undo; Possession Interface README v13 release notes.