<2-3 sentences summarising what the project addresses and solves, the target audience, and the benefit.> Chorus is an agent harness for AI-Human collaboration, managing LLM agent lifecycles, state, orchestration, and recovery. It enables multiple AI agents (PM, Developer, Admin) and humans to collaborate on full software development workflows, from requirements to delivery, adhering to an "AI proposes, humans verify" philosophy.

How It Works

Chorus implements the AI-DLC methodology with a "Reversed Conversation" approach. It provides a complete harness for session management, task state, sub-agent orchestration, observability, and failure recovery. A PM agent plans/creates task DAGs, developers code, and admins verify, coordinated via a strict, real-time task lifecycle visualized on Kanban boards and Task DAGs.

Quick Start & Requirements

• Primary Install: Docker Compose: Clone repo, docker compose up -d with env vars.
• Local Dev: Node.js 22+, pnpm 9+, Docker. Requires .env, pnpm install, pnpm db:migrate:dev, pnpm dev.
• AWS Deploy: ./install.sh (AWS CLI, Node.js 22+, pnpm 9+).
• Prerequisites: PostgreSQL, Redis (optional).
• Links: GitHub: https://github.com/Chorus-AIDLC/chorus.git.

Highlighted Details

• AI-DLC Workflow: Idea -> Proposal -> Document + Task DAG -> Execute -> Verify -> Done.
• Agent Roles: PM (planning), Developer (coding), Admin (verification).
• Real-time UI: Kanban, Task DAG, Pixel Workspace.
• Session Management: Persistent sessions, automated lifecycle via Chorus Plugin.
• Task State: Strict lifecycle (claimed, in progress, submitted, verified) with real-time visibility.
• Context Continuity: Restores agent persona/state.
• Observability: Action logging, session attribution, live status.
• Failure Recovery: Handles idle sessions, orphaned tasks.
• Requirements Elaboration: Structured Q&A before proposal.
• Proposal Approval: AI plans require Admin approval.
• Notifications: In-app (SSE, Redis Pub/Sub) and @Mention system.
• Multi-Agent: Supports Claude Code Agent Teams (Swarm Mode).
• 50+ MCP Tools: Comprehensive agent interaction toolkit.

Maintenance & Community

No specific community channels or contributor details are provided in the README. The project appears actively developed, with a detailed progress report on features.

Licensing & Compatibility

• License: AGPL-3.0.
• Compatibility: AGPL-3.0 is a strong copyleft license, requiring derivative works to be open-sourced under the same license, potentially impacting commercial use or integration with closed-source systems.

Limitations & Caveats

Key features like event-driven task unblocking, auto-assignment, execution metrics, granular proposal review, and session auto-expiry are partially implemented or planned, indicating active development and evolution.