Chorus-AIDLC/Chorus

What it solves

Chorus provides a management layer (a "harness") for teams using multiple AI coding agents. It solves the problem of coordinating AI-driven development by moving beyond simple chat interfaces into a structured pipeline where agents propose changes and humans verify them, ensuring that AI-generated code actually ships as reliable software.

How it works

Chorus implements the AI-Driven Development Lifecycle (AI-DLC) methodology, which uses a "Reversed Conversation" philosophy: AI proposes, and humans verify. It manages the entire session lifecycle, task states, and sub-agent orchestration through a central system.

Key operational components include:

• The Daemon: A runtime that allows local machines to pick up assigned tasks and execute them.
• Permission Matrix: A fine-grained system that controls who (human or agent) can create ideas, write proposals, draft tasks, or verify work.
• Agent Integration: It connects to various agent platforms (like Claude Code, Codex, Kiro, and others) via the MCP SDK.
• Visual Tracking: It provides a Project Resource Graph (mind-map) and Kanban boards to track the live status of ideas, proposals, and tasks.

Who it’s for

It is designed for developers and teams who use AI coding agents and need a structured way to orchestrate multiple agents, maintain observability, and implement human-in-the-loop verification for software production.

Highlights

• Multi-Agent Orchestration: Support for multiple independent agents with their own keys, working directories, and the ability to hand off work via @-mentions.
• Human-in-the-Loop: A strict workflow (Idea → Proposal → Document/Task DAG → Execute → Verify) to ensure human oversight.
• Broad Compatibility: Integrates with a wide range of MCP-compatible agents and platforms.
• Live Observability: Real-time presence indicators and a visual resource graph to see exactly what agents are working on.