Brainstorm Mcp

This server orchestrates debates between multiple AI models to stress test ideas and code. It exposes three main tools: brainstorm for multi-round debates with synthesis verdicts, brainstorm_quick for instant parallel perspectives across models, and brainstorm_review for multi-model code reviews with severity ratings. You can run it in API mode with your own OpenAI, Gemini, DeepSeek, or Ollama keys, or use hosted mode where Claude spawns sub-agents with different models without needing external APIs. Supports debate styles like red team and Socratic questioning, and includes context injection for grounding discussions in actual code or diffs. Useful when you want multiple model opinions before making architectural decisions or merging risky changes.