Intelligence Route

Features

• 3-tier model routing (Haiku, Sonnet, Opus, Agent Booster)
• Learned pattern-based task optimization
• Data-driven agent and model tier selection
• Generates auditable routing rationale
• Records task outcomes for model training

Use Cases

• When initiating any non-trivial task to ensure optimal resource allocation.
• To replace manual agent and model selection with data-driven decisions.
• When transparency in decision-making for task routing is required.
• To continuously improve task routing efficiency through outcome recording.

Non-Goals

• Performing the actual task; it only routes to the appropriate agent.
• Selecting specific tools for the task; it routes to an agent capable of selecting tools.
• Replacing the user's final decision-making authority without clear rationale.

Workflow

1. Receive task description and optional `--why` flag.
2. Call `hooks_route` for agent recommendation.
3. Call `hooks_model-route` for model tier recommendation.
4. Search for similar past patterns using `hooks_intelligence_pattern-search`.
5. Predict outcome confidence using `neural_predict`.
6. Spawn the recommended agent at the recommended model tier.
7. If `--why` was passed, call `hooks_explain` to surface rationale.
8. After task completion, call `hooks_model-outcome` with success status for training.

Practices

• Agent routing
• Model selection
• Task optimization
• Outcome recording
• Decision transparency

Prerequisites

• MCP server with access to specified hooks (`mcp__claude-flow__hooks_*`)
• Task description as input