Consensys

Multi-agent AI code review with debate and voting.

Four AI experts review your code, debate their findings, and vote on the outcome:

• SecurityExpert - Vulnerabilities, injection attacks, security best practices
• PerformanceEngineer - Efficiency, optimization, resource usage
• ArchitectureCritic - Design patterns, SOLID principles, code structure
• PragmaticDev - Maintainability, practicality, real-world tradeoffs

Installation

# From PyPI
pip install consensys

# With web UI support (quotes required for zsh)
pip install 'consensys[web]'

# From source
git clone https://github.com/noah-ing/consensys.git
cd consensys
pip install -e .

Quick Start

# Set your API key
export ANTHROPIC_API_KEY="your-api-key"

# Review a file
consensys review myfile.py

# Quick mode (faster, single round)
consensys review myfile.py --quick

# Review with auto-fix suggestions
consensys review myfile.py --fix

# Stream AI thinking in real-time
consensys review file.py --stream

# CI mode: fail on HIGH severity or above
consensys review file.py --fail-on HIGH

Features

• Multi-agent debate - AI experts discuss and challenge each other's findings
• Consensus voting - Final decision based on expert votes (APPROVE/REJECT/ABSTAIN)
• Smart caching - Avoid redundant API calls for unchanged code
• 14+ languages - Python, JS, TS, Go, Rust, Java, C, C++, Ruby, PHP, Swift, Kotlin, Scala
• CI/CD integration - GitHub Action, pre-commit hooks, fail-on thresholds
• Export options - Markdown and HTML reports
• RedTeam mode - Generate PoC exploits and auto-patches for vulnerabilities
• Prediction market - Agents bet tokens on code quality outcomes
• Code DNA - Extract codebase style patterns and detect anomalies

CLI Reference

# Review a file
consensys review path/to/file.py

# Review inline code snippet
consensys review --code 'def foo(): pass'

# Batch review directory
consensys review-batch src/
consensys review-batch src/ --lang python --parallel 8

# Git integration
consensys diff                  # Review uncommitted changes
consensys commit                # Review staged changes
consensys pr 123                # Review GitHub PR
consensys pr 123 --post         # Post as PR comment

# History
consensys history               # List past sessions
consensys replay abc123         # Replay a review
consensys export abc123 --format md

# Advanced modes
consensys review file.py --redteam    # Generate exploits + patches
consensys review file.py --predict    # Enable prediction market
consensys review file.py --dna        # Check against codebase style
consensys fingerprint src/            # Extract style fingerprint

# Metrics
consensys metrics               # View API usage and costs

Web UI

# Start the web server
consensys web

# Custom host/port
consensys web --host 0.0.0.0 --port 3000

Open http://localhost:8080 in your browser for real-time streaming reviews.

API Endpoints

Endpoint	Method	Description
/api/health	GET	Health check
/api/review	POST	Submit code for review
/api/sessions	GET	List past sessions
/api/sessions/{id}	GET	Get session details
/ws/review	WebSocket	Streaming reviews

curl -X POST http://localhost:8080/api/review \
  -H "Content-Type: application/json" \
  -d '{"code": "def foo(): eval(input())", "language": "python"}'

GitHub Action

Add to .github/workflows/consensys.yml:

name: Code Review

on:
  pull_request:
    types: [opened, synchronize]

jobs:
  review:
    runs-on: ubuntu-latest
    permissions:
      contents: read
      pull-requests: write
    steps:
      - uses: actions/checkout@v4
      - uses: noah-ing/consensys@v1
        with:
          api_key: ${{ secrets.ANTHROPIC_API_KEY }}
          fail_on: 'HIGH'
          min_severity: 'MEDIUM'

Inputs: api_key (required), fail_on, min_severity, quick_mode, post_comment, files

Outputs: decision, issues_count, session_id, summary

Pre-commit Hook

Add to .pre-commit-config.yaml:

repos:
  - repo: https://github.com/noah-ing/consensys
    rev: v0.2.1
    hooks:
      - id: consensys-review           # Quick review
      # - id: consensys-review-strict  # Fail on HIGH
      # - id: consensys-review-all     # All languages

pip install pre-commit
pre-commit install
pre-commit run --all-files

Configuration

Environment Variables

export ANTHROPIC_API_KEY="your-api-key"
export CONSENSYS_MODEL="claude-3-5-haiku-20241022"

Config File

Create .consensys.yaml in your project root:

model: claude-3-5-haiku-20241022
min_severity: MEDIUM
fail_on: HIGH
quick_mode: false
default_team: full-review

Team Presets

consensys set-team --preset security-focused   # SecurityExpert + ArchitectureCritic
consensys set-team --preset quick-check        # SecurityExpert + PragmaticDev
consensys set-team SecurityExpert PragmaticDev # Custom team

Python API

from consensys import DebateOrchestrator
from consensys.personas import PERSONAS

orchestrator = DebateOrchestrator(personas=PERSONAS)
result = orchestrator.run_full_debate(code, context="User input handler")

print(result.final_decision)  # APPROVE, REJECT, or ABSTAIN
print(result.key_issues)

Docker

# Build and run
docker build -t consensys .
docker run -p 8080:8080 -e ANTHROPIC_API_KEY=your-key consensys

Docker Compose:

version: "3.8"
services:
  consensys:
    build: .
    ports:
      - "8080:8080"
    environment:
      - ANTHROPIC_API_KEY=${ANTHROPIC_API_KEY}
    volumes:
      - consensys-data:/app/data
volumes:
  consensys-data:

RedTeam Mode

Generate proof-of-concept exploits and auto-patches for security vulnerabilities:

consensys review vulnerable.py --redteam

Supports SQL injection, XSS, command injection, path traversal, and auth bypass. All exploits are marked for authorized testing only.

Prediction Market

Agents bet tokens on code quality predictions. Track accuracy over time:

consensys review file.py --predict
consensys predict list
consensys predict resolve abc123 --outcome safe
consensys predict leaderboard

Code DNA

Extract codebase style patterns and detect anomalies in new code:

consensys fingerprint src/           # Extract patterns
consensys review file.py --dna       # Check against patterns

Detects naming convention violations, missing type hints, docstring drift, import style differences, and copy-paste indicators.

Examples

The examples/ directory contains:

• vulnerable.py - 12 common security vulnerabilities
• clean.py - Secure alternatives and best practices
• review-demo.sh - CLI usage patterns
• .consensys.yaml - Example configuration

# Try it
consensys review examples/vulnerable.py --quick
consensys review examples/clean.py --quick

VS Code Extension

See vscode-extension/README.md for the VS Code integration.

License

MIT

Contributing

Contributions welcome! Open an issue or submit a PR.

---

Built with Claude by Anthropic