githubnext/chopin

What it solves

Chopin addresses the challenge of keeping technical documentation, such as RFCs, implementation plans, and decision records, synchronized with the actual state of a software repository. It provides a collaborative environment where human developers and AI agents can co-author documents that are grounded in the repository's current code and pull requests.

How it works

Chopin creates "channels" that link a rich-text document (backed by MDX) with a shared chat interface. Users can interact with an AI agent, the "Planner," by tagging it in chat or using commands within the document. The Planner uses read-only tools to inspect the repository, propose changes, and draft content. It tracks decisions separately from the prose, ensuring that team consensus is preserved even if the document is later edited. It also supports integration with external coding agents via the Model Context Protocol (MCP).

Who it’s for

This project is for software engineering teams who want to bridge the gap between their planning documentation and their codebase, allowing AI to assist in maintaining technical specifications and decision logs.

Highlights

• Multiplayer Collaboration: Real-time document editing with presence indicators and cursors for both humans and AI.
• Repository-Grounded AI: The Planner agent can read repository files, history, and pull requests to inform its drafting and planning.
• Decision Tracking: Maintains a record of questionnaire answers and comments, preventing important team decisions from being lost during document revisions.
• Extensible Architecture: Supports connecting external coding agents via MCP to bridge the gap between planning and implementation.
• Research Workflow: Allows users to trigger automated research tasks that generate child documents, keeping complex investigations organized.