This project transforms the `oracle-models` skill into a **local MCP Server**. It classifies development tasks (LIGHT, MEDIUM, HEAVY) based on their technical complexity and suggests the most cost-efficient AI model using aggregated data from multiple AI leaderboard sources.
Classifies development task complexity (LIGHT/MEDIUM/HEAVY) and suggests the most cost-efficient AI model per provider — works as both an **MCP server** (live tools) and an **agent skill** (behavioral instructions). Data sourced from the [Artificial Analysis Intelligence Index](https://artificialanalysis.ai/leaderboards/models).