Model Selection

You can route LLM agent tasks by setting a persistent model override in a local JSON configuration file. The system uses a hierarchical resolution mechanism to automatically select the optimal Claude model for each agent spawn.

Hierarchical model selection evaluates a 5-layer priority system to resolve the optimal LLM for agent spawns. It checks user preferences, agent charters, and task-type heuristics to automatically choose between high-reasoning and cost-effective models.

Yes, you can set a persistent default model by instructing the system to always use a specific LLM. This preference is saved in a local JSON configuration file, ensuring the override applies consistently across all future sessions.

If a preferred LLM model is unavailable, the system maintains operational continuity by implementing a robust fallback chain. This ensures that agent spawns continue processing tasks without interruption using alternative models.

You can manage LLM costs by configuring task-aware defaults that route complex coding tasks to high-reasoning models like Claude Sonnet, while routing routine documentation tasks to cost-effective models like Claude Haiku.

No external dependencies are required to configure LLM routing. The system manages model selection and persistent overrides entirely through a local JSON configuration file using built-in task-type heuristics.