Multi-MCP

This server orchestrates multiple AI models (OpenAI, Anthropic, Google, OpenRouter) to run parallel code analysis through four main tools: codereview for systematic OWASP Top 10 security checks, compare for side-by-side model outputs on architectural questions, debate for multi-agent consensus with critique rounds, and chat for context-aware development assistance. It runs as an MCP stdio server that Claude Code or OpenCode connects to, handling async execution so three models finish in roughly 10 seconds instead of 30. Configuration lives in YAML files where you can add custom models or CLI-based tools like Ollama, with aliases letting you request "mini" instead of "gpt-5-mini". Reach for it when you want second opinions from multiple LLMs on code quality, security flaws, or design decisions without manually querying each API.