cerebustech-dev/rahho-mcp

rahho-mcp

Multi-agent code review MCP server for Claude Code.

Status: ✅ Fully tested and operational ()

Quick Start

Just want to get started? See for a 5-minute setup guide.

Documentation:

• - Get up and running in 5 minutes
• - Complete installation and configuration
• - How to test your installation
• - Validation and performance metrics

Overview

rahho-mcp is a Model Context Protocol (MCP) server that provides multi-vendor AI code review capabilities for Claude Code. Get independent code reviews from multiple AI providers in a single tool call.

Why Multi-Agent Review?

Different AI models have different strengths:

• Google Gemini 2.5 Pro: Excellent at performance analysis and large codebase understanding
• OpenAI GPT-4.1: Strong security focus and threat modeling
• Local Inference (GB10): Free, private edge case analysis using Qwen2.5-Coder

By combining multiple perspectives, you get more comprehensive code review coverage.

Features

• Multi-vendor review: Gemini, GPT-4, and local inference in one call
• Security-focused: Dedicated security auditing with GPT-4.1
• Cost-effective: Local inference for edge cases (free)
• MCP native: Purpose-built for Claude Code integration
• Async-first: Fast, non-blocking execution
• Structured output: Rich review results with cost tracking

Installation

# Clone the repository
git clone https://github.com/cerebustech-dev/rahho-mcp.git
cd rahho-mcp

# Install dependencies
pip install -e .

# For development
pip install -e ".[dev]"

Configuration

Add to your Claude Code MCP configuration (.claude/mcp.json):

{
  "mcpServers": {
    "rahho-mcp": {
      "command": "python",
      "args": ["-m", "rahho_mcp.server"],
      "cwd": "/path/to/rahho-mcp",
      "env": {
        "ANTHROPIC_API_KEY": "${ANTHROPIC_API_KEY}",
        "OPENAI_API_KEY": "${OPENAI_API_KEY}",
        "GOOGLE_API_KEY": "${GOOGLE_API_KEY}",
        "GB10_API_BASE": "http://localhost:30000/v1"
      }
    }
  }
}

Usage

Once configured, Claude Code can invoke multi-agent review:

# Claude Code automatically calls this when you ask for multi-vendor review
result = multi_agent_review(
    code="your code here",
    context="what this code does",
    focus_areas=["security", "performance", "edge_cases"]
)

Example Review Flow

User: "Implement AWS Secrets Manager rotation with security review"

Claude Code: [implements code]

Claude Code: Let me get multi-vendor security review...
<calls rahho-mcp tool>

Results:
- Gemini 2.5 Pro: Found performance issue in connection pooling
- GPT-4.1: Identified missing IAM permission and race condition
- GB10 Qwen: Caught edge case with partial rotation failure

Claude Code: [applies fixes based on all three reviews]

Architecture

┌─────────────────────────────────────┐
│  Claude Code (User Interface)       │
├─────────────────────────────────────┤
│  MCP Server (rahho-mcp)             │
│  ├─ ReviewOrchestrator              │
│  ├─ Gemini Agent                    │
│  ├─ OpenAI Agent                    │
│  └─ Local Inference Agent           │
└─────────────────────────────────────┘

Relationship to rahho

This project is inspired by and extracted from rahho, a CLI-based multi-agent orchestration tool. While rahho focuses on workflow orchestration via CLI and JSON manifests, rahho-mcp is purpose-built as an MCP server for real-time code review in Claude Code.

Development

# Run tests
pytest

# Type checking
mypy src/

# Code formatting
black src/ tests/
ruff check src/ tests/

License

Apache License 2.0

Author

Rod Davenport