LearnNewsExamplesServices
Frontmatter
id8288
titleNeo Agent OS: Orchestration & Swarm Architecture
stateOpen
labels
epicstaleaiarchitecture
assignees[]
createdAtJan 3, 2026, 9:03 AM
updatedAtApr 4, 2026, 5:44 AM
githubUrlhttps://github.com/neomjs/neo/issues/8288
authortobiu
commentsCount1
parentIssuenull
subIssues[]
subIssuesCompleted0
subIssuesTotal0
blockedBy[]
blocking[]

Neo Agent OS: Orchestration & Swarm Architecture

Openepicstaleaiarchitecture
tobiu
tobiu commented on Jan 3, 2026, 9:03 AM

This epic tracks the architectural exploration and implementation of the Neo Agent OS, a Node.js-based runtime for autonomous AI agent swarms.

Core Concepts:

  1. The Orchestrator: A persistent Node.js loop that manages the lifecycle of autonomous agents (PM, Dev, QA) and replaces the human-in-the-loop for standard tasks.
  2. Context Graph: A structured state model (Nodes/Edges) representing the project reality (Tickets, Files, Concepts, Agents), allowing agents to share a "Shared Brain" and avoid context window overflows.
  3. Decision Trees: The "Standard Operating Procedures" (SOPs) or algorithms that define legal transitions and fallback logic for agents (e.g., "If tests fail 3 times, escalate").
  4. Swarm Intelligence: The collective behavior emerging from specialized agents (e.g., "Night Watchman" for releases, "Bug Hunter" for regressions) working via the Context Graph.

Architecture:

  • Runtime: Node.js (Persistent, distinct from the transient Browser App).
  • Bridge: Neural Link (WSS) for controlling Neo.mjs Apps.
  • Memory: Neo Memory Core for long-term state and session summaries.
  • Interface: MCP for tool access.

Goals:

  • Define the supervisor.mjs prototype (The "Night Watchman").
  • Design the JSON schema for the Context Graph.
  • Implement the first "Decision Node" logic (e.g., simplistic Release automation).
tobiu added the epic label on Jan 3, 2026, 9:03 AM
tobiu added the ai label on Jan 3, 2026, 9:03 AM
tobiu added the architecture label on Jan 3, 2026, 9:03 AM