githubnext/chopin

What it solves

Chopin addresses the challenge of creating durable technical documentation—such as implementation plans, RFCs, and decision records—through a collaborative process between human developers and AI agents. It prevents the loss of decision context by separating the reasoning and accepted answers from the final prose, ensuring that later rewrites do not silently erase the history of why a choice was made.

How it works

Chopin creates "channels" that link a GitHub repository to a multiplayer rich-text document. It uses a hosted agent called Planner that can read the repository's code and pull requests to propose changes and draft content.

Key components include:

• Collaborative Document: A shared MDX-backed space where humans and the agent edit together in real-time.
• Chat: A team communication space where the agent can be summoned using @chopin to act on the conversation.
• Decisions: A dedicated system to track attributed questionnaire answers and comments separately from the document text.
• Research: A /research command that triggers background jobs to gather information and publish the results as child documents.
• MCP Integration: An Model Context Protocol (MCP) endpoint that allows external coding agents to create or consume implementation plans.

Who it’s for

Software engineering teams who need to co-author technical specifications and planning documents with AI assistance while maintaining a strict record of architectural decisions.

Highlights

• Repository-Grounded AI: The Planner agent has read-only access to the GitHub repository to ensure proposals are based on actual code.
• Decision Persistence: Separates the "why" (decisions) from the "what" (the document prose).
• Multiplayer Editing: Supports real-time co-authoring with presence and cursors for both humans and agents.
• External Agent Connectivity: Uses MCP to bridge the gap between high-level planning in Chopin and actual code implementation by external agents.