Agent skill

brainstorm

Run multi-round AI brainstorming debates between multiple LLM providers (GPT, DeepSeek, Groq, Ollama). Claude actively participates as a debater alongside external models. Use when the user wants diverse perspectives, multi-model critiques, or synthesized answers from several AI models working together.

Stars 54
Forks 2

Install this agent skill to your Project

npx add-skill https://github.com/spranab/brainstorm-mcp/tree/main/.claude/skills/brainstorm-mcp

Metadata

Additional technical details for this skill

author
spranab
version
1.1.0

SKILL.md

Brainstorm — Multi-Model AI Debates

Use the brainstorm-mcp tools to orchestrate structured debates between multiple LLMs. By default, Claude participates as an active debater alongside external models — reading their responses, pushing back, building on ideas, and refining its position across rounds.

When to Use

  • User says "brainstorm this", "get multiple perspectives", "debate this topic"
  • A question benefits from diverse viewpoints rather than a single model's answer
  • User wants to compare how different models approach a problem
  • Architecture decisions, trade-off analysis, or open-ended design questions

Tools

Tool Description
brainstorm Run a multi-round debate between configured AI models
brainstorm_respond Submit Claude's response for the current round of an interactive session
list_providers Show all configured providers, models, and API key status
add_provider Dynamically add a new AI provider at runtime

Core Workflow (Interactive Mode — Default)

1. Start the debate

brainstorm({ topic: "Best architecture for a real-time app", rounds: 3 })

Returns round 1 external model responses + a session_id.

2. Respond each round

Read the external models' responses, form your own position, then call:

brainstorm_respond({ session_id: "<id>", response: "Your substantive contribution..." })

This stores your response and runs the next external round. Repeat until all rounds complete.

3. Synthesis

After the final round response, synthesis runs automatically and returns the full debate.

Non-Interactive Mode

For debates between external models only (no Claude participation):

brainstorm({ topic: "React vs Vue", participate: false })

Parameters

brainstorm

Parameter Type Default Description
topic string required What to brainstorm about
models string[] all providers Specific models as provider:model
rounds number 3 Number of debate rounds (1-10)
synthesizer string first model Model for final synthesis
systemPrompt string Custom system prompt for all models
participate boolean true Whether Claude joins as an active debater

brainstorm_respond

Parameter Type Description
session_id string Session ID from the brainstorm tool
response string Claude's contribution (min 50 chars)

Best Practices

  • Use 2-3 rounds for quick opinions, 4-5 for deeper analysis
  • Specify models explicitly when you want particular perspectives
  • Use systemPrompt to focus the debate on specific aspects
  • Check list_providers first to see which models are available
  • One model failing won't abort the debate — results are resilient
  • Engage with other models' specific points — agree, disagree, build upon, or challenge

Expand your agent's capabilities with these related and highly-rated skills.

davila7/claude-code-templates

verl-rl-training

Provides guidance for training LLMs with reinforcement learning using verl (Volcano Engine RL). Use when implementing RLHF, GRPO, PPO, or other RL algorithms for LLM post-training at scale with flexible infrastructure backends.

23,776 2,298
Explore
davila7/claude-code-templates

openrlhf-training

High-performance RLHF framework with Ray+vLLM acceleration. Use for PPO, GRPO, RLOO, DPO training of large models (7B-70B+). Built on Ray, vLLM, ZeRO-3. 2× faster than DeepSpeedChat with distributed architecture and GPU resource sharing.

23,776 2,298
Explore
davila7/claude-code-templates

gguf-quantization

GGUF format and llama.cpp quantization for efficient CPU/GPU inference. Use when deploying models on consumer hardware, Apple Silicon, or when needing flexible quantization from 2-8 bit without GPU requirements.

23,776 2,298
Explore
davila7/claude-code-templates

Claude Code Guide

Master guide for using Claude Code effectively. Includes configuration templates, prompting strategies "Thinking" keywords, debugging techniques, and best practices for interacting with the agent.

23,776 2,298
Explore
davila7/claude-code-templates

qdrant-vector-search

High-performance vector similarity search engine for RAG and semantic search. Use when building production RAG systems requiring fast nearest neighbor search, hybrid search with filtering, or scalable vector storage with Rust-powered performance.

23,776 2,298
Explore
davila7/claude-code-templates

behavioral-modes

AI operational modes (brainstorm, implement, debug, review, teach, ship, orchestrate). Use to adapt behavior based on task type.

23,776 2,298
Explore

Didn't find tool you were looking for?

Be as detailed as possible for better results