Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
What This Skill Does
Mixture of Agents (MoA) sends a prompt to 3 frontier models in parallel, then passes all three responses to a fourth model that synthesizes them into a single answer. Costs ~$0.03 per query on the paid tier using OpenRouter.
Running three models in parallel and synthesizing their outputs catches blind spots that any single model, including GPT-4 or Claude Opus, would miss, at a lower total cost than a single Opus call.
When to use it
- Evaluating startup competitive moats before an investment decision
- Comparing technical approaches like RLHF vs DPO for alignment
- Cross-referencing factual claims across models with different training data
- Generating a market landscape overview with diverse model perspectives
- Getting a second and third opinion on a high-stakes architectural decision
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: Analyze the competitive landscape for AI code generation — who has defensible moats?
- 1Send the prompt simultaneously to Kimi K2.5, GLM-5, and MiniMax M2.5
- 2Wait for all three proposer responses (parallel, ~45s)
- 3Pass all three responses to the aggregator model (Kimi K2.5)
- 4Aggregator resolves conflicts and synthesizes the strongest insights
A structured analysis combining unique perspectives from all three models, with conflicting claims resolved and the most defensible points surfaced
Requirements
Accounts, API keys, or tools you or your AI assistant may need to set up while using this skill.