bayze-mcp automates cross-model review in Cherry Studio
The Juejin AI article author announced bayze-mcp, an MIT-licensed MCP server that lets Cherry Studio assistants coordinate drafting, review, and revision across model providers through one tool call. The report provides installation and YAML configuration instructions, while its example uses DeepSeek for drafting, Qwen Plus for review, and Doubao Lite for revision. The author also claims low costs, 8–15-second latency, local key storage, and no telemetry, but these performance and privacy claims lack independent verification.
The project offers a concrete way for Cherry Studio users to automate cross-provider model review, although the reported workflow configuration conflicts with the described revision model.
你有没有遇到过这种情况？
你在用 AI 助理（Cherry Studio、Claude Desktop、Cursor）写一段重要内容，想让「另一个模型的 AI 帮忙检查一下」——但只能手动复制粘贴，在 ChatGPT、Claude、DeepSeek 之间来回切换。
bayze-mcp 把这个流程自动化了：一个 MCP 工具调用，触发「起草 → 评审 → 修订」的跨厂商协作。
效果演示
$ bayze run --workflow review_then_revise_cn \
--input "用一句话解释 HTTP 和 HTTPS 的区别"
✓ draft deepseek-chat 2.1s ~$0.001
✓ review qwen-plus 2.7s ~$0.001
verdict: revise — "缺少 TLS 握手和非对称加密"
✓ revise doubao-lite 3.6s
→ 修订稿采纳了评审的纠正，技术准确性提升
三家国产模型协作：DeepSeek 写初稿、通义千问评审（结构化 JSON 结论）、豆包按需修订。每一步都可审计：哪个模型、多少 token、什么费用。