superai-mcp

✨ Features

🔀 Multi-Model Broadcast

Send the same prompt to Claude, Codex, and Gemini in parallel and get aggregated results in a single response. Perfect for multi-perspective code reviews, comparing model reasoning, or finding consensus across different AI engines.

• Per-target model and parameter overrides via models and overrides
• System prompt injection for consistent instructions across all targets
• Built-in prompt templates: review, refactor, explain, test, debug, optimize

🔗 Chain Pipeline

Build sequential multi-model pipelines where each step's output automatically flows into the next. Chain different models for tasks that benefit from staged processing — draft with one, refine with another, verify with a third.

• Output auto-injected as <previous_output> tags between steps
• Stops on first failure with partial results returned
• End-to-end timeout budget across all steps

🗳️ Vote Consensus

Run candidates in parallel, then have a judge model pick the best answer. Get the most reliable output by leveraging the strengths of multiple models simultaneously.

• Judge auto-excluded from candidate pool
• Only successful candidate outputs forwarded to judge
• Fully configurable candidate and judge selection

💬 Multi-Round Debate

Stage an alternating debate between two models over multiple rounds. Each round sees the opponent's previous response. Ideal for exploring trade-offs, stress-testing solutions, or generating balanced analysis.

• Configurable number of debate rounds
• Opponent output injected as <opponent_response> tags
• Any two CLIs as debate sides