chorus-codes/chorus

What it solves

AI coding tools often produce subtle bugs or exhibit biases that the same model cannot detect in its own work. Chorus solves this by orchestrating multiple different AI vendors (such as Claude, GPT, and Gemini) to review the same code change in parallel, ensuring that a "second opinion" from a different model catches blind spots before code is shipped.

How it works

Chorus runs as a local daemon and web UI (the "Cockpit") that manages AI tools as subprocesses on your machine. It leverages your existing AI subscriptions (Claude Pro, ChatGPT Plus, Gemini Advanced) by driving their CLIs headlessly, meaning there is no additional per-token cost for these users. It uses a Model Context Protocol (MCP) server to integrate with IDEs and other AI assistants. The system uses templates to define which models act as the writer and which act as reviewers, requiring a specific quorum of agreement (e.g., unanimous) to green-light a change.

Who it’s for

Developers who use AI to write code and want a high-confidence verification layer without paying for expensive API keys or sending their code to a new third-party vendor.

Highlights

• Multi-Vendor Review: Mixes models from different lineages (Anthropic, OpenAI, Google, etc.) to avoid shared training biases.
• Subscription-Based: Uses existing AI CLI tools to keep out-of-pocket costs at $0 for subscription users.
• Local-First: Runs locally on the laptop; code only reaches the AI vendors the user already trusts.
• Persona System: Reviewers can be assigned specific roles (e.g., Sentinel for security, Profiler for performance) to focus their critique.
• MCP Integration: Can be triggered directly from tools like Claude Code, Cursor, or Windsurf.