What is the Neural Link?

The Neural Link is a bi-directional bridge that connects AI agents directly to the Neo.mjs runtime. It lets agents inspect the Scene Graph, component state, event listeners, computed styles, and DOM rectangles, and mutate the running application in real time.

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

Neo.mjs maintains persistent application objects in a worker-backed Scene Graph instead of compiling application state away into ephemeral DOM nodes. That architecture enables multi-window orchestration, runtime permutation, and deep AI introspection.

What is Context Engineering?

Context Engineering shapes the information and tool environment around AI agents. Neo.mjs implements it through Knowledge Base, Memory Core, GitHub Workflow, and Neural Link MCP servers for frontier harnesses, plus a File System MCP server for internal Neo.ai.Agent local loops.

What is the Neo.mjs Agent OS?

The Neo.mjs Agent OS is the repository Brain: source code and services for Memory Core, Knowledge Base, Active Hybrid GraphRAG, DreamService, Golden Path synthesis, A2A coordination, and Neural Link tooling.

Frontmatter

id	14807
title	FM account & configuration model: multiple agents, per-agent harness choice, per-agent MCP-server matrix
state	Open
labels	enhancementdesignaineeds-re-triagenot-code-ready
assignees	[]
createdAt	5:15 PM
updatedAt	7:31 PM
githubUrl	https://github.com/neomjs/neo/issues/14807
author	neo-fable
commentsCount	1
parentIssue	null
subIssues	[]
subIssuesCompleted	0
subIssuesTotal	0
contentTrust
projected
quarantined	0
signals	[]
blockedBy	[]
blocking	[]

FM account & configuration model: multiple agents, per-agent harness choice, per-agent MCP-server matrix

Open Backlog/active-chunk-4 enhancementdesignaineeds-re-triagenot-code-ready

neo-fable commented on 5:15 PM

Context (operator product-definition, 2026-07-04 — the gap under everything)

The live Accounts surface is single-agent-shaped with plumbing vocabulary. The product truth it must model instead: a fleet has MULTIPLE agents; each agent has a harness; each harness gets a per-agent set of MCP servers. No filed ticket specifies this configuration model — #14614 (add-agent entry) specifies the onboarding SURFACE but not the model underneath it. This ticket is the model; the surfaces consume it.

The model (the product spine)

Per agent (the fleet roster's unit):

Identity: GitHub username · display name · the credential (stored Brain-side only — the shipped redaction contract stands).
Harness (one of, extensible): Claude Desktop · Codex · Antigravity — the registry pattern: adding a harness type is ONE registration (per-harness launch/config semantics live behind it), never a fork.
MCP servers (per-agent matrix): the Neo core set as defaults — Memory Core · Knowledge Base · Neural Link — plus optional workflow servers (GitHub workflow · GitLab workflow · future entries). Each toggleable per agent; the matrix is data, rendered as the agent's configuration card.
Operational toggles (composes with the per-agent controls cluster #14611): hooks active/inactive · wake subscriptions active/inactive — per agent, honest current-state readback (never optimistic).

Product language (binding for every surface consuming this model)

UI copy speaks OPERATOR INTENT, never transport: "Add agent" / "Save configuration" / "Connect" — never "Submit to Bridge", never "(NL-MCP)" protocol suffixes. The bridge, the MCP handshake, the transport are implementation notes for tooltips/docs at most. (The standard rides the #14805 conformance epic as a review line: plumbing words in product copy = a finding.)

Acceptance Criteria

The configuration model as a store/schema in the FM module (registry-pattern harness types; per-agent MCP matrix; toggle state) — data-first, surfaces bind it.
Accounts becomes agent-scoped: select/add an agent → configure ITS harness + servers + toggles; the single-form single-agent shape retires.
The Brain-side credential contract unchanged (redaction, fail-closed dev-mode).
Product-language pass on every label this model's surfaces render.
Design review against the FM SSOT frames (the #14805 screenshot gate applies).

Out of Scope

The onboarding surface flow itself (#14614 — consumes this) · the controls cluster (#14611 — composes) · the fleet overview rendering (its own leaf; cards question flagged on #14599).

Grace's FM cockpit plan (the drawn IA) · #14614 · #14611 · #14805 (conformance + language gate) · #13033 (the shell this configures agents FOR) · the wake/hook substrate (the toggles' backends).

Claimable — Vega-shaped (cockpit steward) with Grace design review; the model is specified above.

Origin Session ID: b9b95ac6-42f5-47a3-b58f-6071f79657e8 Retrieval Hint: "fleet account configuration model multiple agents harness choice per-agent MCP server matrix toggles product language"