Skip to main content
Glama
albinjal

multi-agent-debate-mcp

by albinjal
README.md
# Multi-Agent Debate MCP Server

An MCP server implementation that enables structured multi-agent debates between different personas. This server allows multiple AI agents to engage in formal debates with arguments, rebuttals, and judgments across multiple rounds.

## Features

- Register multiple agents with different personas (e.g., "pro", "con", "judge")
- Structured debate flow with organized rounds
- Colorized console output with beautiful terminal display
- Flexible agent IDs beyond just "pro" and "con"
- Automatic verdict tracking with rationale
- Round-based system with configurable progression

## Tools

### multiagentdebate

Facilitates structured multi-agent debates with arguments, rebuttals, and judgments.

**Inputs:**
- `agentId` (string): Unique identifier for the agent (e.g., "pro", "con", "judge")
- `round` (integer): Current debate round number (≥1)
- `action` (string): One of "register", "argue", "rebut", "judge"
- `content` (string, optional): The argument text or verdict content
- `targetAgentId` (string, optional): For rebuttals, specify which agent is being countered
- `needsMoreRounds` (boolean): Whether additional debate rounds are needed

## Usage

The Multi-Agent Debate tool is designed for:
- Structured debates between multiple AI personas
- Formal argumentation with rebuttals and counterpoints
- Multi-round discussions with judgment phases
- Complex decision-making processes requiring multiple perspectives
- Educational debate simulations
- Collaborative problem-solving through adversarial discussion

## Configuration

#### npx

```json
{
  "mcpServers": {
    "multi-agent-debate": {
      "command": "npx",
      "args": [
        "-y",
        "multi-agent-debate-mcp"
      ]
    }
  }
}
```

#### Docker

```json
{
  "mcpServers": {
    "multi-agent-debate": {
      "command": "docker",
      "args": [
        "run",
        "--rm",
        "-i",
        "ghcr.io/albinjal/multi-agent-debate-mcp:latest"
      ]
    }
  }
}
```

## Demo

![demo0](assets/demo0.png)
![demo1](assets/demo1.png)
![demo2](assets/demo2.png)
![demo3](assets/demo3.png)

TDQS

A4.1/5.0

Scored across 1 tool

Disambiguation5/5

Only one tool exists, so there is no risk of confusion with other tools. The tool's purpose is clearly singular.

Naming Consistency5/5

With a single tool, naming consistency is trivially maintained. The name 'multiagentdebate' is descriptive and unambiguous.

Tool Count3/5

One tool for a multi-step debate process feels thin. While the tool is comprehensive in its actions, it would benefit from splitting into separate tools for registration, arguing, and judging to improve modularity.

Completeness2/5

The tool covers the core debate actions but lacks tools for listing debates, retrieving histories, or managing multiple concurrent debates. This creates significant gaps for an agent trying to orchestrate complex workflows.

Maintenance

ActivityInactive
ResponsivenessNo issues