Verified

Deepseek MCP Server

DeepSeek V4 API access including chat, completion, and account balance tools.

Data last scanned today · Reviewed today

537 tools★ 351TypeScriptMIT

Overview

This server implements the Model Context Protocol to provide direct access to the DeepSeek V4 API surface. It allows your AI environment to interact with specific endpoints for chat completions, FIM completions, and account management, supporting features like structured output, reasoning effort controls, and native function tools. The implementation maintains a narrow scope by exposing only the documented API surfaces. It avoids additional abstractions, providing raw access to models like deepseek-v4-flash and deepseek-v4-pro while managing conversation memory and streaming responses as defined by the current API documentation.

Our verdict

With a security score of 43/100, this server is classified as high-risk. This rating is driven primarily by the fact that the authentication methods have not yet been reviewed and the server is community-maintained rather than developed by the official vendor. Users should exercise caution and review the codebase before deployment, as these factors leave the security posture unverified.

Supports DeepSeek V4 chat, FIM, and model listing functionality.
Small footprint with only three direct dependencies.
Maintained via an active GitHub repository with 351 stars.
Provides documented access to account balance and API availability.

Authentication mechanisms remain unreviewed by security auditors.
Read-only mode and safety boundaries are not editorially verified.
Community-maintained status lacks official vendor support or guarantees.

Setup

Claude Desktop

{
  "mcpServers": {
    "deepseek-mcp-server": {
      "args": [
        "-y",
        "deepseek-mcp-server"
      ],
      "command": "npx"
    }
  }
}

Paste into ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows), then restart Claude Desktop. Some servers require API keys or tokens as environment variables — check the project's README and add an "env" object if needed.

No read-only mode — running this grants full read/write access.
No built-in authentication — anyone with access to this config can invoke every tool it exposes.

Tools

Tool	Description	Risk
chat_completion	DeepSeek V4 chat.	low
create_response	Stateless native Responses API calls with text or structured input, reasoning effort controls, function tools, server-side web search, structured output, and semantic streaming.	low
completion	DeepSeek V4 Pro FIM completion.	low
list_models	Reads the live DeepSeek model list.	low
get_user_balance	Reads account balance and availability.	low
reset_conversation	Clears an in-memory conversation.	medium
list_conversations	Lists in-memory conversation IDs.	low

Compatibility

Client	Local	Docker	Remote	Read-only
ChatGPT
Claude Desktop
Cursor
VS Code
Windsurf

Frequently asked questions

›What authentication is required to use this server?

You must provide a DEEPSEEK_API_KEY environment variable. If using the hosted remote instance, you also need to provide a Bearer token via the Authorization header.

›What models are supported?

The server supports the DeepSeek V4 API, including deepseek-v4-flash and deepseek-v4-pro models.

›How do I handle CORS for browser-based MCP clients?

You can configure the MCP_HTTP_ALLOWED_ORIGINS environment variable with a comma-separated list of allowed origins. Requests with an Origin header that are not on this list will be rejected with a 403 error.

›Does this server store conversation history?

Yes, it supports in-memory conversation memory, which can be managed using the list_conversations and reset_conversation tools. You can also limit the history size using the CONVERSATION_MAX_MESSAGES environment variable.