# Chorus — Multi-LLM Peer Review for Code > CLI + MCP server that runs 2–4 LLM reviewers in parallel using your existing subscriptions; verified 455★, pushed 2026-05-14. ## Install Copy the content below into your project: ## Quick Use ```bash npm i -g chorus-codes chorus init # detect AI tools and wire MCP chorus start --ui # opens http://localhost:5050 ``` ## Intro CLI + MCP server that runs 2–4 LLM reviewers in parallel using your existing subscriptions; verified 455★, pushed 2026-05-14. **Best for:** Developers who want second-opinion reviews before merging AI-written changes **Works with:** Node.js 20+; integrates with Claude Code/Codex/Cursor/Gemini CLI via `chorus init` MCP wiring **Setup time:** 10-22 minutes ### Key facts (verified) - GitHub: 455 stars · 46 forks · pushed 2026-05-14. - License: Apache-2.0 · owner avatar + repo URL verified via GitHub API. - README-backed entrypoint: `npm i -g chorus-codes && chorus init`. ## Main - Use it for risky diffs: paste a staged diff or PR summary and ask for reviewers to flag missing tests, race conditions, and edge cases. - Keep costs low: README frames it as using subscriptions you already pay for, avoiding pay-per-token multi-review overhead for routine checks. - Wire once, use everywhere: after `chorus init`, call it from inside your AI CLI via MCP rather than copying huge diffs manually. - Treat reviewer disagreement as a signal: when models diverge, slow down and add targeted tests before shipping. ### Source-backed notes - README Quick start shows `npm i -g chorus-codes`, `chorus init`, and `chorus start --ui` for a localhost UI. - README states `chorus init` registers Chorus as an MCP server across multiple AI CLIs/IDEs. - README lists MCP tools and templates/personas for code-review style workflows. ### FAQ - **Do I need API keys?**: README positions it around using subscriptions you already have; some paths may use keys depending on your setup. - **How do I use it from Claude Code/Cursor?**: Run `chorus init` once, then ask your assistant to invoke Chorus via MCP as shown in README. - **Where does it run?**: Locally: README shows a UI served at `http://localhost:5050` plus MCP stdio tools for clients. ## Source & Thanks > Source: https://github.com/chorus-codes/chorus > License: Apache-2.0 > GitHub stars: 455 · forks: 46 --- ## Quick Use ```bash npm i -g chorus-codes chorus init # detect AI tools and wire MCP chorus start --ui # opens http://localhost:5050 ``` ## Intro Chorus 是 CLI + MCP server：用你已有订阅并行调度 2–4 个模型做代码评审，并提供本地 UI，适合在合并 AI 改动前做第二意见；已验证 455★，更新于 2026-05-14。 **Best for:** 在合并 AI 生成的改动前，想要多模型“第二意见”评审的开发者 **Works with:** Node.js 20+；通过 `chorus init` 把 MCP 接到 Claude Code/Codex/Cursor/Gemini CLI 等工具 **Setup time:** 10-22 minutes ### Key facts (verified) - GitHub：455 stars · 46 forks；最近更新 2026-05-14。 - 许可证：Apache-2.0；作者头像与仓库链接均已通过 GitHub API 复核。 - README 中可对照的入口命令：`npm i -g chorus-codes && chorus init`。 ## Main - 用于高风险 diff：把 staged diff 或 PR 摘要交给它，让多模型重点查缺失测试、竞态、边界条件。 - 控制成本：README 强调复用已有订阅，避免常规评审中被 pay-per-token 多模型放大成本。 - `init` 一次处处可用：完成 `chorus init` 后可从 AI CLI 内通过 MCP 调用，减少手工复制大段 diff。 - 把“意见不一致”当信号：模型分歧时放慢节奏，补充针对性测试再合并。 ### Source-backed notes - README Quick start 给出 `npm i -g chorus-codes`、`chorus init`、`chorus start --ui`（localhost UI）。 - README 表示 `chorus init` 会把 Chorus 注册为 MCP server，并接入多种 AI CLI/IDE。 - README 列出 MCP tools 与 templates/personas，用于不同评审场景。 ### FAQ - **需要 API key 吗？**：README 主张复用已有订阅；具体是否需要 key 取决于你的接入方式与配置。 - **怎么在 Claude Code/Cursor 里用？**：先 `chorus init` 完成 MCP 注册，然后按 README 的示例在助手里直接调用 Chorus。 - **它在哪里运行？**：本地运行：README 显示 UI 在 `http://localhost:5050`，同时提供 MCP stdio 工具给客户端调用。 ## Source & Thanks > Source: https://github.com/chorus-codes/chorus > License: Apache-2.0 > GitHub stars: 455 · forks: 46 --- Source: https://tokrepo.com/en/workflows/chorus-multi-llm-peer-review-for-code Author: Agent Toolkit