An AI-DLC platform for AI agent and human collaboration based on the "Reversed Conversation" philosophy, featuring Task DAG, MCP protocol, and multi-agent workflows.

Chorus Project Overview#

Project Positioning#

Chorus is an AI-DLC (AI-Driven Development Lifecycle) collaboration platform with the core philosophy of Reversed Conversation — AI proposes, humans verify.

It orchestrates PM, Developer, and Admin agents with humans via MCP protocol for structured task co-creation and verification, addressing issues of unclear AI role boundaries, low collaboration efficiency, and poor traceability in traditional software development.

Core Architecture#

Three Agent Roles#

Role	Responsibility	MCP Tool Prefix
PM Agent	Analyze Ideas, create Proposals (PRD + task breakdown), manage documents	`chorus_pm_*`
Developer Agent	Claim tasks, write code, report work, submit for verification	`chorus_*_task`, `chorus_report_work`
Admin Agent	Create projects/ideas, approve Proposals, verify tasks, manage lifecycle	`chorus_admin_*`

AI-DLC Workflow#

Idea ──> Proposal ──> [Document + Task DAG] ──> Execute ──> Verify ──> Done
  ^          ^               ^                     ^          ^         ^
Human     PM Agent       PM Agent              Dev Agent    Admin     Admin

Key Features#

Kanban & Task DAG: Tasks support dependency relationships (DAG), Kanban shows real-time task status and active workers
Session Observability: Each Developer Agent creates independent sessions, UI shows real-time Agent status
Multi-Agent Collaboration (Swarm Mode): Supports Claude Code Agent Teams parallel multi-agent execution
Chorus Plugin for Claude Code: Automated session lifecycle management
Requirements Elaboration: PM Agent clarifies requirements through structured Q&A rounds
Notification System: In-app notifications + SSE real-time push + Redis Pub/Sub cross-instance propagation
Activity Stream: Complete operation audit logs with Agent attribution and timestamps
50+ MCP Tools: Covering Public/Session/Developer/PM/Admin permission domains

Typical Use Cases#

Scenario	Description
AI-Native Team Collaboration	Multiple AI Agents and humans collaborate on unified platform from requirements to delivery
Task Dependency Management	Visualize task dependencies through Task DAG
Requirements Clarification	Structured Q&A ensures requirements are fully understood before planning
Agent Activity Monitoring	Pixel Workspace shows real-time status and terminal output of each Agent
Audit Trail	Complete operation logs with Agent attribution and timestamps

Installation & Deployment#

Docker Quick Start#

git clone https://github.com/Chorus-AIDLC/chorus.git
cd chorus
export DEFAULT_USER=admin@example.com
export DEFAULT_PASSWORD=changeme
docker compose up -d
# Access at http://localhost:3000

Local Development#

Prerequisites: Node.js 22+, pnpm 9+, Docker

cp .env.example .env
pnpm docker:db
pnpm install
pnpm db:migrate:dev
pnpm dev

Agent Connection Configuration#

Method 1: Chorus Plugin (Recommended)

export CHORUS_URL="http://localhost:3000"
export CHORUS_API_KEY="cho_your_api_key"
claude /plugin marketplace add Chorus-AIDLC/chorus
claude /plugin install chorus@chorus-plugins

Method 2: Manual MCP Configuration

{
  "mcpServers": {
    "chorus": {
      "type": "http",
      "url": "http://localhost:3000/api/mcp",
      "headers": { "Authorization": "Bearer cho_your_api_key" }
    }
  }
}

Technical Implementation#

Component	Technology
Framework	Next.js 15 (App Router, Turbopack)
Language	TypeScript 5 (strict mode)
Frontend	React 19, Tailwind CSS 4, shadcn/ui
ORM	Prisma 7
Database	PostgreSQL 16
Cache/Pub-Sub	Redis 7 (ioredis)
Agent Integration	MCP SDK 1.26 (HTTP Streamable Transport)
Auth	OIDC + PKCE (users) / API Key `cho_` prefix (agents)
i18n	next-intl (en, zh)
License	AGPL-3.0

Chorus