Consensus Loop

v1.0.0Coding Agentsstable

Claude Code plugin — structural guardrails for multi-agent software development. RTM-based evidence, cross-model adversarial audit, 7 MCP tools, worktree isolation, HITL retrospective. DOI: 10.5281/zenodo.19108370

ai-agentsaudit-historyclaude-codeclaude-code-plugincode-review
Share:
14
Stars
0
Downloads
0
Weekly
0/5

What is Consensus Loop?

Consensus Loop is a Model Context Protocol (MCP) server that allows AI assistants like Claude, Cursor, and VS Code to claude code plugin — structural guardrails for multi-agent software development. rtm-based evidence, cross-model adversarial audit, 7 mcp tools, worktree isolation, hitl retrospective. doi: 10.5281/ze...

Claude Code plugin — structural guardrails for multi-agent software development. RTM-based evidence, cross-model adversarial audit, 7 MCP tools, worktree isolation, HITL retrospective. DOI: 10.5281/zenodo.19108370

This server falls under the Coding Agents category on MCPgee, the world's largest MCP server directory with 33,000+ servers.

Features

  • Claude Code plugin — structural guardrails for multi-agent s

Use Cases

Implement structural guardrails for multi-agent development teams.
Run cross-model adversarial audits for code quality.
Enable human-in-the-loop review for agent-generated code.
berrzebb

Maintainer

LicenseMIT
Languagejavascript
Versionv1.0.0
UpdatedApr 29, 2026
Statushealthy
Maintenanceactive

Works with

ClaudeOpenAIwindowsmacoslinux

Installation

Manual Installation

npx consensus-loop

Configuration

Configuration Details

Config File

claude_desktop_config.json

Performance

Response Metrics

Response Time< 200ms
ThroughputMedium

Resource Usage

Memory UsageLow
CPU UsageLow

How to Set Up and Use Consensus Loop

Consensus Loop is a Claude Code plugin that adds structural guardrails to multi-agent software development by providing RTM-based evidence tracking, cross-model adversarial audits, and a human-in-the-loop retrospective workflow. It exposes 7 deterministic MCP tools for code mapping, dependency graph analysis, pattern scanning, coverage measurement, and audit history queries, all integrated into Claude Code's PostToolUse lifecycle. Development teams use it to enforce quality standards across AI-generated code by requiring structured approval before changes are merged.

Prerequisites

  • Claude Code installed and configured (the plugin targets the Claude Code CLI)
  • Node.js LTS for running the plugin infrastructure
  • Access to a second AI model (e.g., GPT-4) for cross-model adversarial audits
  • vitest set up with JSON coverage output if you want coverage_map tool support
  • An MCP-compatible client — Claude Code is the primary target
1

Add the plugin marketplace source

Register the berrzebb plugin marketplace in Claude Code so it can discover the consensus-loop plugin.

claude plugin marketplace add berrzebb/claude-plugins
2

Install the consensus-loop plugin

Install consensus-loop from the registered marketplace source.

claude plugin install consensus-loop@berrzebb-plugins
3

Copy the example configuration to your project

Create the .claude/consensus-loop directory in your project and copy the example config and templates.

cp ~/.claude/plugins/cache/berrzebb-plugins/consensus-loop/*/examples/config.example.json \
   .claude/consensus-loop/config.json
cp -r ~/.claude/plugins/cache/berrzebb-plugins/consensus-loop/*/examples/en/templates/ \
      .claude/consensus-loop/templates/
4

Edit config.json for your project

Configure the review watch file path and the trigger, agree, and pending tags to match your team's workflow.

{
  "plugin": { "locale": "en" },
  "consensus": {
    "watch_file": "docs/review/author.md",
    "trigger_tag": "[REVIEW_NEEDED]",
    "agree_tag": "[APPROVED]",
    "pending_tag": "[CHANGES_REQUESTED]"
  }
}
5

Start a work session with the orchestrator

Use the orchestrator slash command to begin task distribution and engage the consensus loop.

/consensus-loop:orchestrator
6

Trigger a cross-model adversarial audit

Run the manual audit command to send the current code diff to a second AI model for adversarial review.

/consensus-audit

Consensus Loop Examples

Client configuration

The plugin is installed via Claude Code's plugin system, not via a direct MCP JSON block. After installation the MCP tools are available automatically in Claude Code sessions.

{
  "mcpServers": {
    "consensus-loop": {
      "command": "npx",
      "args": ["consensus-loop"],
      "env": {}
    }
  }
}

Prompts to try

Example slash commands and prompts once the plugin is active in Claude Code.

- "/consensus-loop:planner — start 6-phase requirement definition for a new feature"
- "/consensus-loop:verify — run all 8 done-criteria checks on current changes"
- "/consensus-status — show the current loop state and pending reviews"
- "Use code_map to find all exported functions in the src/auth directory"
- "Run audit_scan on the checkout module to flag hardcoded strings and type-safety issues"

Troubleshooting Consensus Loop

Plugin commands are not recognized after installation

Restart Claude Code after installation. Confirm the plugin installed correctly by checking ~/.claude/plugins/cache/berrzebb-plugins/consensus-loop/ exists.

coverage_map tool returns empty results

The tool reads vitest JSON coverage output. Ensure you have run vitest with the --coverage flag and that the output file path matches what the plugin expects.

Cross-model audit step hangs or fails

The adversarial audit requires credentials for a second model provider (e.g., an OpenAI API key). Verify that the model configuration in config.json points to a reachable endpoint with valid credentials.

Frequently Asked Questions about Consensus Loop

What is Consensus Loop?

Consensus Loop is a Model Context Protocol (MCP) server that claude code plugin — structural guardrails for multi-agent software development. rtm-based evidence, cross-model adversarial audit, 7 mcp tools, worktree isolation, hitl retrospective. doi: 10.5281/zenodo.19108370 It connects AI assistants to external tools and data sources through a standardized interface.

How do I install Consensus Loop?

Follow the installation instructions on the Consensus Loop GitHub repository. Clone the repo, install dependencies, and add the server config to your AI client.

Which AI clients work with Consensus Loop?

Consensus Loop works with all major MCP-compatible AI clients including Claude Desktop, Claude Code, Cursor, VS Code (GitHub Copilot), Windsurf, and Cline.

Is Consensus Loop free to use?

Yes, Consensus Loop is open source and available under the MIT license. You can use it freely in both personal and commercial projects.

Browse More Coding Agents MCP Servers

Explore all coding agents servers available in the MCPgee directory. Each server includes setup guides for Claude, Cursor, and VS Code.

Quick Config Preview

{ "mcpServers": { "consensus-loop": { "command": "npx", "args": ["-y", "consensus-loop"] } } }

Add this to your claude_desktop_config.json or .cursor/mcp.json

Read the full setup guide →

Ready to use Consensus Loop?

Browse our complete directory of 33,000+ MCP servers, read setup guides for your editor, and start building with the Model Context Protocol.

33,000+ ServersFree & Open SourceStep-by-Step Guides