Skip to content

CLI LLM Council Guide: Collaborative Multi-Agent Decisions

Run the LLM Council directly from command line for collaborative decision-making with multiple AI agents through peer review and synthesis.

Overview

The LLM Council creates a collaborative environment where:

  1. Multiple Perspectives: Each council member (GPT-5.1, Gemini, Claude, Grok) independently responds
  2. Peer Review: Members evaluate and rank each other's anonymized responses
  3. Synthesis: A Chairman synthesizes the best elements into a final answer

Basic Usage

Step 1: Run a Simple Query

swarms llm-council --task "What are the best practices for code review?"

Step 2: Enable Verbose Output

swarms llm-council --task "How should we approach microservices architecture?" --verbose

Step 3: Process the Results

The council returns: - Individual member responses - Peer review rankings - Synthesized final answer


Use Case Examples

Strategic Business Decisions

swarms llm-council --task "Should our SaaS startup prioritize product-led growth or sales-led growth? Consider market size, CAC, and scalability."

Technology Evaluation

swarms llm-council --task "Compare Kubernetes vs Docker Swarm for a startup with 10 microservices. Consider cost, complexity, and scalability."

Investment Analysis

swarms llm-council --task "Evaluate investment opportunities in AI infrastructure companies. Consider market size, competition, and growth potential."

Policy Analysis

swarms llm-council --task "What are the implications of implementing AI regulation similar to the EU AI Act in the United States?"

Research Questions

swarms llm-council --task "What are the most promising approaches to achieving AGI? Evaluate different research paradigms."

Council Members

The default council includes:

Member Model Strengths
GPT-5.1 Councilor gpt-5.1 Analytical, comprehensive
Gemini 3 Pro Councilor gemini-3-pro Concise, well-processed
Claude Sonnet 4.5 Councilor claude-sonnet-4.5 Thoughtful, balanced
Grok-4 Councilor grok-4 Creative, innovative
Chairman gpt-5.1 Synthesizes final answer

Workflow Visualization

┌─────────────────────────────────────────────────────────────────┐
│                        User Query                                │
└─────────────────────────────────────────────────────────────────┘
┌─────────────┬─────────────┬─────────────┬─────────────┐
│   GPT-5.1   │  Gemini 3   │ Claude 4.5  │   Grok-4    │
│  Councilor  │  Councilor  │  Councilor  │  Councilor  │
└─────────────┴─────────────┴─────────────┴─────────────┘
┌─────────────────────────────────────────────────────────────────┐
│                   Anonymized Peer Review                         │
│         Each member ranks all responses (anonymized)             │
└─────────────────────────────────────────────────────────────────┘
┌─────────────────────────────────────────────────────────────────┐
│                        Chairman                                  │
│         Synthesizes best elements from all responses             │
└─────────────────────────────────────────────────────────────────┘
┌─────────────────────────────────────────────────────────────────┐
│                     Final Synthesized Answer                     │
└─────────────────────────────────────────────────────────────────┘

Best Practices

Query Formulation

  • Be specific and detailed in your queries
  • Include context and constraints
  • Ask for specific types of analysis

When to Use LLM Council

  • Complex decisions requiring multiple perspectives
  • Research questions needing comprehensive analysis
  • Strategic planning and evaluation
  • Questions with trade-offs to consider

Performance Tips

  • Use --verbose for detailed progress tracking
  • Expect responses to take 30-60 seconds
  • Complex queries may take longer

Limitations

  • Requires multiple API calls (higher cost)
  • Not suitable for simple factual queries
  • Response time is longer than single-agent queries

Command Reference

swarms llm-council --task "<query>" [--verbose]
Option Type Default Description
--task string Required Query for the council
--verbose flag False Enable detailed output

Next Steps