🛡️ AEGIS Router

Advanced Engineered Governance & Intelligence System

AEGIS is the neural routing cortex for your OpenClaw Agent Swarm. It analyzes every prompt across 15 cognitive dimensions in real-time and routes it to the optimal model tier from your local infrastructure — all in under 1ms with zero external API calls.

Philosophy

"The most powerful intelligence isn't the one that knows the most — it's the one that knows which intelligence to summon."

AEGIS embodies the principle of intelligent delegation. In a world where dozens of AI models exist with different strengths, costs, and specializations, blindly routing every request to the most expensive model is wasteful. AEGIS acts as the strategic command layer — a triage system that ensures:

• Simple greetings don't waste premium compute
• Code generation goes to models optimized for structured output
• Deep reasoning tasks activate the most capable minds in your arsenal
• Cost is optimized without sacrificing quality

Execution Tiers

Tier	Name	Models	Use Case
🟢 T1	EFFICIENT	Gemini 3 Flash	Greetings, quick lookups, heartbeats
🔵 T2	BALANCED	Qwen3 Coder, Llama 4 Scout	Code generation, technical Q&A
🟠 T3	ADVANCED	Claude 4.5 Opus	Complex analysis, architecture design
🔴 T4	REASONING	Kimi K2.5	Multi-step reasoning, proofs, deep logic

Integration

As OpenClaw Middleware

AEGIS is designed to intercept messages before the agent processes them:

import { createAegis } from '@disier/aegis-router';

const aegis = createAegis();

// In your OpenClaw gateway middleware:
function onMessage(context) {
  const result = aegis.intercept({
    prompt: context.message,
    sessionId: context.session.id,
    agentId: context.agent.id,
  });

  // Apply the routing decision
  context.session.model = result.decision.model;
  
  console.log(result.decision.reason);
}

Standalone Usage

import { createAegis } from '@disier/aegis-router';

const aegis = createAegis();

// Analyze a prompt
const result = aegis.intercept({
  prompt: "Prove that the halting problem is undecidable using a diagonal argument",
});

// result.decision.tier → REASONING
// result.decision.model → "kimi-k2.5"
// result.decision.confidence → 0.97

Configuration

Edit aegis.config.yaml to customize:

• Model mappings — Point tiers to your available models
• Scoring weights — Tune the 15 dimensions to your workload
• Thresholds — Adjust where tier boundaries fall
• Session pinning — Enable/disable model consistency per conversation
• Auto-escalation — Control reasoning marker escalation

Swarm Compatibility

AEGIS is designed to work within an OpenClaw Agent Swarm. In a multi-agent setup, each agent can have its own model override in the swarm config. AEGIS will respect agent-level overrides while still performing analysis (for logging and metrics).

Pass agentId and optional modelOverride in the context:

const result = aegis.intercept({
  prompt: userMessage,
  sessionId: session.id,
  agentId: 'code-specialist',
  modelOverride: 'qwen3-coder', // Hard lock for this agent
});