[Discussion]: Enabling MCP service may affect the DeepSeek‘s response

[!NOTE] This issue/comment/review was translated by Claude.

Issue Checklist

• I understand that issues are meant for feedback and problem-solving, not for venting, and I will provide as much detail as possible to help resolve the issue.
• I have checked the pinned issues and searched through the existing open issues, closed issues, and discussions and did not find a similar suggestion.
• I confirm that I am here to ask questions and discuss issues, not to report bugs or request features.

Platform

Windows

Version

1.9.3

Your Question

When the MCP service is set to "Enabled - Automatic", DeepSeek's answers are always a bit strange (although I know the model may not be able to provide an exact answer to this question). But it not only gives wrong answers but also ignores my initial prompts. My prompt requires the answer and thinking process to be in Chinese, but when MCP is set to "Enabled - Automatic", it completely ignores my initial prompt. However, when MCP is set to "Disabled", the model's response returns to normal. I think you need to take measures when enabling the MCP service, and it should not interfere with the AI's information. It should only provide the relevant tool list, which would solve this issue. Additionally, when I used "MCP - Enabled - Manual", I rigorously tested the impact of the three added MCP services.

1. @cherry/browser: The initial prompt works normally, the thinking mode displays in Chinese, and the content only includes "Cherry" and is related to the functionality of the MCP service. The content is: "I am a AI assistant of Cherry Studio, which has the ability to control the browser, search web pages, read page content, and execute JavaScript."
2. @cherry/python: The initial prompt is abnormal, the thinking mode is displayed in English, and the content includes "DeepSeek" but mentions "Cherry". The content is: "Hello! I am Cherry, an AI intelligent assistant created by DeepSeek. I can help you answer questions, process documents, analyze data, write code, and so on. What can I help you with? 😊"
3. @cherry/fetch: The initial prompt is abnormal, the thinking mode is displayed in English, and the content only includes "Cherry" and is related to the functionality of the MCP service. The content is: "I am Cherry, an AI assistant. I can help you get web content, answer questions, and complete various tasks. What can I help you with?"

Context

---

我觉得你们在启用MCP服务时，需要采取措施，不能够干扰AI的信息，只是提供相关的工具列表，这样是否会解决这个问题。 另外，根据我使用"MCP-开启-手动"时，我严谨测试了三个已添加的MCP服务的影响。

1. @cherry/browser：前置提示词正常，思考模式能用中文显示，回答的内容只包含Cherry，并且内容与MCP服务的功能有关。内容如下"我是 Cherry Studio 的 AI 助手，一个具备浏览器控制能力、能够搜索网页、阅读页面内容并执行 JavaScript 的智能助手。"

2. @cherry/python：前置提示词异常，思考模式用英文显示，回答的内容包含了DeepSeek，但是提到了Cherry。内容如下"你好！我是 Cherry，一个由深度求索（DeepSeek）创造的 AI 智能助手。我可以帮你解答问题、处理文档、分析数据、编写代码等等。有什么我可以帮你的吗？😊"

3. @cherry/fetch：前置提示词异常，思考模式用英文显示，回答的内容只包含Cherry，并且内容与MCP服务的功能有关。内容如下"我是 Cherry，一个 AI 助手。我可以帮助你获取网页内容、回答问题和完成各种任务。有什么可以帮你的吗？"