Skip to content
MCP AI Agent Guidelines

routing-adapt

Advancedworkflow

Mission

Section titled “Mission”

Deploy → observe → reinforce → prune → converge. Workflows that get smarter over time.

When to Use

Section titled “When to Use”

ONLY use when an existing multi-agent workflow needs autonomous bio-inspired route optimization based on historical performance.

Do NOT use for: general research, design, review, debugging, planning, implementation, code quality, or documentation. If unsure, use the specific domain tool instead.

Triggers: “bio-inspired routing”, “adaptive workflow”, “reinforce good routes”, “optimize routing history”, “ant colony”, “Hebbian learning”

Skills Invoked

Section titled “Skills Invoked”
  • adapt-aco-router — ant colony pheromone trail reinforcement
  • adapt-annealing — simulated annealing to escape local optima
  • adapt-hebbian-router — Hebbian learning for co-skill chain optimization
  • adapt-physarum-router — slime mould efficient path finding
  • adapt-quorum — quorum-threshold-triggered escalation

Chain-To

Section titled “Chain-To”
  • agent-orchestrate — integrate adaptive routing into orchestration
  • quality-evaluate — measure routing quality improvement over time

Safety Requirements

Section titled “Safety Requirements”

Always pair with:

  1. resil-membrane isolation between routing stages
  2. gov-policy-validation gate before production deployment
  3. Monitoring via qm-observable-extractor if ENABLE_PHYSICS_SKILLS=true

Example

Section titled “Example”
{
  "request": "Optimize routing in our 5-agent code review pipeline using historical success data",
  "options": { "algorithm": "hebbian", "sessions": 50 }
}

Output: Learned routing weights, pruned low-performing paths, recommended configuration update for orchestration.toml.