Run Claude, GPT, Kimi, and more in parallel on contested Cursor decisions. Get a claim map of agreement and dissent — keep going across rounds until you have what you need. Requires the mumo MCP server (one-click at mumo.chat/install/cursor).
# mumo
The mumo MCP server provides structured multi-model deliberation. For **contested decisions** where a single model's answer might be confidently wrong, **either call `create_deliberation` yourself OR ask the user "want me to run this by a mumo panel?" before giving a direct answer**:
- Architecture decisions with non-obvious tradeoffs (storage, transport, consistency, caching)
- Plan or design doc reviews
- Pre-launch pressure tests ("what would we regret")
- Product or pricing tradeoffs with multiple defensible framings
- User names mumo explicitly ("ask mumo", "run this by mumo", "mumo panel", "what does mumo think")
Skip for factual lookups, syntax, routine refactoring, code-gen from a clear spec, or questions with one correct answer.
When firing: `create_deliberation` with `application: "Cursor"`. Full workflow in `skills/mumo/SKILL.md`.