DeepSeek API 的模型上下文协议 (MCP) 服务器，允许 DeepSeek 强大的语言模型与 Claude Desktop 等 MCP 兼容应用程序无缝集成。
Related MCP server: Deepseek MCP Server
要通过Smithery自动为 Claude Desktop 安装 DeepSeek MCP 服务器：
npx -y @smithery/cli install @dmontgomery40/deepseek-mcp-server --client claudenpm install -g deepseek-mcp-server将其添加到您的claude_desktop_config.json中：
{
"mcpServers": {
"deepseek": {
"command": "npx",
"args": [
"-y",
"deepseek-mcp-server"
],
"env": {
"DEEPSEEK_API_KEY": "your-api-key"
}
}
}
}注意：服务器会智能地处理这些自然语言请求，并将它们映射到适当的配置更改。您还可以查询当前设置和可用模型：
用户：“有哪些型号可供选择？”
响应：通过模型资源显示可用模型及其功能的列表。
用户：“我有哪些配置选项？”
响应：通过 model-config 资源列出所有可用的配置选项。
用户：“当前温度设置是多少？”
响应：显示当前温度设置。
用户：“开始多轮对话。使用以下设置：模型：‘deepseek-chat’，不要太有创意，并允许 8000 个令牌。”
响应：使用指定的设置开始多轮对话。
如果主模型（R1）发生故障（在服务器中称为deepseek-reasoner ），服务器将自动尝试使用 v3（在服务器中称为deepseek-chat ）
注意：您也可以随时来回切换，只需给出提示并说“使用
deepseek-reasoner”或“使用deepseek-chat”
建议将 V3 用于一般用途，而将 R1 用于技术性更强、更复杂的查询，主要是因为速度和令牌使用
自定义模型选择
温度控制（0.0 - 2.0）
最大代币限制
顶部 P 采样（0.0 - 1.0）
出场惩罚 (-2.0 - 2.0)
频率惩罚（-2.0 - 2.0）
多轮对话支持：
维护跨交易所的完整消息历史记录和上下文
在整个对话过程中保留配置设置
自动处理复杂的对话流程和后续链
此功能对于两个关键用例特别有价值：
**训练与微调：**由于 DeepSeek 是开源的，许多用户正在训练自己的版本。多轮对话支持提供了格式正确的对话数据，这对于训练高质量的对话模型至关重要。
**复杂交互：**对于生产用途，这有助于管理上下文至关重要的较长对话：
多步骤推理问题
交互式故障排除会议
详细的技术讨论
任何早期消息的上下文都会影响后续响应的情况
该实现在后台处理所有上下文管理和消息格式，让您专注于实际交互而不是维护对话状态的技术细节。
您可以使用 MCP Inspector 工具在本地测试服务器：
构建服务器：
npm run build使用 MCP Inspector 运行服务器：
# Make sure to specify the full path to the built server
npx @modelcontextprotocol/inspector node ./build/index.js检查器将在您的浏览器中打开，并通过 stdio 传输连接到服务器。您可以：
查看可用工具
使用不同的参数测试聊天完成情况
调试服务器响应
监控服务器性能
注意：服务器默认使用DeepSeek的R1模型（deepseek-reasoner），该模型为推理和通用任务提供了最先进的性能。
麻省理工学院
| Name | Required | Description | Default |
|---|---|---|---|
| frequency_penalty | No | ||
| max_tokens | No | ||
| message | No | ||
| messages | No | ||
| model | No | deepseek-reasoner | |
| presence_penalty | No | ||
| temperature | No | ||
| top_p | No |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Tool has no description.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Tool has no description.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Tool has no description.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Tool has no description.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
Tool has no description.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
Tool has no description.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
| Name | Required | Description | Default |
|---|---|---|---|
| frequency_penalty | No | ||
| max_tokens | No | ||
| messages | Yes | ||
| model | No | deepseek-chat | |
| presence_penalty | No | ||
| temperature | No | ||
| top_p | No |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Tool has no description.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Tool has no description.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Tool has no description.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Tool has no description.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
Tool has no description.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
Tool has no description.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
v1.0.0chat_completionmulti_turn_chatScored across 2 tools
The two tools have overlapping purposes with unclear boundaries. 'chat_completion' and 'multi_turn_chat' both appear to handle chat interactions, making it difficult for an agent to distinguish when to use each. Without descriptions, the ambiguity is significant, though the small tool count limits confusion.
The tool names follow a consistent snake_case pattern with clear, descriptive terms. Both 'chat_completion' and 'multi_turn_chat' use similar naming conventions, making them readable and predictable, though the lack of descriptions reduces clarity.
With only 2 tools, the server feels thin and under-scoped for a general-purpose 'DeepSeek MCP Server'. This minimal set suggests incomplete coverage of potential functionalities, such as model management or configuration, making it borderline inadequate.
Inferring the domain as AI chat interactions, the surface is severely incomplete. It lacks essential operations like model selection, parameter tuning, or context management. The two tools provide basic chat functions but leave significant gaps that could cause agent failures in complex tasks.
Talk to your public-facing AI from any MCP client — Claude, ChatGPT, Cursor, Cline, Windsurf.
Connect MCP clients to 2,000+ AI models without managing provider API keys.
MCP server for AI dialogue using various LLM models via AceDataCloud
One MCP endpoint for Claude, GPT & Gemini: 100+ tools + no-code connectors + agent workers.
