LLMForge - Autonomous AI Engineering, LLM Evals, Fine-Tuning & Inference Cost Engine MCP

Universal AI engineering engine for Cursor, Claude Code & Windsurf. Architects complete LLM blueprints (prompting vs RAG vs fine-tuning vs agents matrix, model selection with token cost projections), builds eval harnesses (golden datasets, LLM-as-judge with bias controls, CI regression gates), produces fine-tuning recipes (LoRA/QLoRA configs with concrete hyperparameters, honest should-you-finetune gates, distillation paths), slashes inference bills 50-90% (cascade routing, prompt caching, batch