Skip to main content
Glama

DeepSeek MCP 服务器

适用于当前 DeepSeek V4 API 的模型上下文协议 (MCP) 服务器。

截至 2026 年 4 月 24 日,DeepSeek 公共 API 参考文档包含:

  • POST /chat/completions,支持 deepseek-v4-flash 和 deepseek-v4-pro

  • POST /beta/completions,用于 V4 Pro FIM 补全

  • GET /models

  • GET /user/balance

此服务器仅公开上述已记录的 API 接口。它不提供 V4 监控、推测性图像/视频/上传工具或自动模型替换功能。

工具

  • chat_completion: DeepSeek V4 聊天。默认为 deepseek-v4-flash。支持 thinking: { "type": "enabled" | "disabled" }、reasoning_effort: "high" | "max"、JSON 输出、函数工具、logprobs、流式传输和对话记忆。

  • completion: DeepSeek V4 Pro FIM 补全。默认为 deepseek-v4-pro。

  • list_models: 读取实时 DeepSeek 模型列表。

  • get_user_balance: 读取账户余额和可用性。

  • reset_conversation: 清除内存中的对话。

  • list_conversations: 列出内存中的对话 ID。

Related MCP server: Deepseek MCP Server

远程托管

  • URL: https://deepseek-mcp.ragweld.com/mcp

  • 认证: Authorization: Bearer <token>

Codex CLI:

export DEEPSEEK_MCP_AUTH_TOKEN="REPLACE_WITH_TOKEN"
codex mcp add deepseek --url https://deepseek-mcp.ragweld.com/mcp --bearer-token-env-var DEEPSEEK_MCP_AUTH_TOKEN

Claude Code:

export DEEPSEEK_MCP_AUTH_TOKEN="REPLACE_WITH_TOKEN"
claude mcp add --transport http deepseek https://deepseek-mcp.ragweld.com/mcp --header "Authorization: Bearer $DEEPSEEK_MCP_AUTH_TOKEN"

Cursor:

node -e 'const fs=require("fs"),p=process.env.HOME+"/.cursor/mcp.json";let j={mcpServers:{}};try{j=JSON.parse(fs.readFileSync(p,"utf8"))}catch{};j.mcpServers={...(j.mcpServers||{}),deepseek:{url:"https://deepseek-mcp.ragweld.com/mcp",headers:{Authorization:"Bearer ${env:DEEPSEEK_MCP_AUTH_TOKEN}"}}};fs.mkdirSync(process.env.HOME+"/.cursor",{recursive:true});fs.writeFileSync(p,JSON.stringify(j,null,2));'

本地 Stdio

DEEPSEEK_API_KEY="REPLACE_WITH_DEEPSEEK_KEY" npx -y deepseek-mcp-server

Docker:

docker pull docker.io/dmontgomery40/deepseek-mcp-server:0.5.0
docker run --rm -i -e DEEPSEEK_API_KEY="REPLACE_WITH_DEEPSEEK_KEY" docker.io/dmontgomery40/deepseek-mcp-server:0.5.0

环境变量

必需:

DEEPSEEK_API_KEY=your-api-key

可选:

DEEPSEEK_BASE_URL=https://api.deepseek.com
DEEPSEEK_REQUEST_TIMEOUT_MS=120000
DEEPSEEK_DEFAULT_MODEL=deepseek-v4-flash
MCP_TRANSPORT=stdio
MCP_HTTP_HOST=127.0.0.1
MCP_HTTP_PORT=3001
MCP_HTTP_PATH=/mcp
MCP_HTTP_STATEFUL_SESSION=false
CONVERSATION_MAX_MESSAGES=200

验证

npm run build
npm test
DEEPSEEK_API_KEY="REPLACE_WITH_DEEPSEEK_KEY" npm run test:live
DEEPSEEK_MCP_AUTH_TOKEN="REPLACE_WITH_TOKEN" npm run test:remote

实时冒烟测试会执行真实的 DeepSeek 请求,包括模型列表、余额、非思考模式聊天、带有 reasoning_content 的思考模式流式聊天、FIM 补全以及 MCP 工具调用。

注册表标识

  • MCP 注册表名称: io.github.DMontgomery40/deepseek

  • npm 软件包: deepseek-mcp-server

  • OCI 软件包: docker.io/dmontgomery40/deepseek-mcp-server:0.5.0

官方参考

许可证

MIT

Available Tools

11 tools
chat_completionA

Primary DeepSeek V4.1 chat tool for text, images, and multi-turn generation. Defaults to deepseek-flash. Provide either message (a text string or text/image/file content parts) or messages (full chat history); if both are provided, messages is used. Images may use an HTTP(S) URL, supported base64 data URL, or Files API file_id. Thinking is enabled by default; use reasoning_effort for current effort controls. Use conversation_id for explicit in-memory context. Set include_raw_response=true only for provider-payload debugging.

Inputs: stop, model, tools, top_p, +19 more. Returns: model, usage, status, message, +10 more

completionA

DeepSeek FIM completion tool for prompt/suffix fill-in-the-middle workflows. Defaults to deepseek-flash in non-thinking mode. Use this when you need raw completion text instead of chat message formatting. Set include_raw_response=true only when you need the full provider payload for debugging.

Inputs: prompt (required), echo, stop, model, +10 more. Returns: text, model, usage, status, +7 more

create_responseA

Create a stateless DeepSeek V4.1 response using the OpenAI-compatible Responses API. Defaults to deepseek-flash and accepts text, image URLs, base64 images, and Files API file_id inputs. Use reasoning.effort for thinking control, tools for function or server-side web-search tools, and stream=true for semantic SSE aggregation. Send full input history for multi-turn work. Set include_raw_response=true only for provider-payload debugging.

Inputs: text, user, input, model, +11 more. Returns: error, model, usage, status, +10 more

delete_fileB
DestructiveIdempotent

Delete one uploaded DeepSeek image by file_id. This permanently removes that provider-side file reference.

Inputs: file_id (required). Returns: id, object, status, deleted, +4 more

get_user_balanceA
Read-onlyIdempotent

Return the current DeepSeek account balance and availability status. This tool takes no parameters and is read-only. Use it for account health checks when diagnosing provider-side failures.

Inputs: no named parameters. Returns: status, message, retryable, error_type, +3 more

list_conversationsA
Read-onlyIdempotent

List all conversation IDs currently stored in this MCP process memory. This tool takes no parameters and does not call the DeepSeek API. Useful for debugging conversation persistence behavior.

Inputs: no named parameters. Returns: count, status, message, retryable, +3 more

list_filesA
Read-onlyIdempotent

List images uploaded to the DeepSeek Files API. Supports cursor pagination, creation-time order, and the user_data purpose filter.

Inputs: after, limit, order, purpose. Returns: data, object, status, last_id, +6 more

list_modelsA
Read-onlyIdempotent

List available DeepSeek models for model selection and validation. This tool takes no parameters. Use it before passing an explicit model ID to generation tools.

Inputs: no named parameters. Returns: data, object, status, message, +3 more

reset_conversationA
DestructiveIdempotent

Delete stored in-memory chat history for a conversation_id. Use this when you want to keep the same ID but start a fresh thread. This only affects server-side memory in the current MCP process.

Inputs: conversation_id (required). Returns: status, message, removed, retryable, +3 more

retrieve_fileB
Read-onlyIdempotent

Retrieve metadata for one DeepSeek Files API image by file_id.

Inputs: file_id (required). Returns: id, bytes, object, status, +8 more

upload_fileA

Upload a JPEG, PNG, GIF, or WebP image to DeepSeek and return a reusable file_id. Supply base64 image data directly; server-local file paths are intentionally not accepted. Optional expiry is 3,600 to 2,592,000 seconds.

Inputs: filename (required), file_data (required), expires_after_seconds. Returns: id, bytes, object, status, +8 more

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A3.9/5.0

    Scored across 11 tools

    Disambiguation4/5

    The three generation tools (chat_completion, completion, create_response) have overlapping capabilities like text/image input and multi-turn support, which could cause confusion. However, their descriptions clearly differentiate use cases (chat vs FIM vs stateless Responses API), and all other tools target distinct resources or actions.

    Naming Consistency4/5

    Nine of eleven tools follow a consistent verb_noun pattern (e.g., list_models, upload_file, delete_file). The two exceptions (chat_completion and completion) are noun-only names, but they are conventional and readable, making the set mostly predictable.

    Tool Count5/5

    The server provides 11 tools, which is well within the ideal 3–15 range. Each tool covers a distinct aspect of the DeepSeek API (generation, file management, account info, conversation memory), with no obvious redundancy or bloat.

    Completeness4/5

    The surface covers core generation, file CRUD (upload/list/retrieve/delete), model listing, balance check, and conversation memory management. Minor gaps include no explicit update or detailed inspection of conversations, but these are not critical for typical workflows.

    Maintenance

    ActivityMaintained
    ResponsivenessLimited data

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      A
      quality
      A
      maintenance
      MCP server for DeepSeek AI models (Chat + Reasoner). Supports multi-turn sessions, model fallback with circuit breaker, function calling, thinking mode, JSON output, multimodal input, and cost tracking.
      3
      456 npm
      26
      MIT
    • A
      license
      A
      quality
      B
      maintenance
      Bridges MCP clients to the DeepSeek API for chat completions and model discovery, with an optional locked-down bridge that delegates prompts to local CLI harnesses like Claude, Codex, or opencode.
      4
      46 npm
      MIT
    • A
      license
      Not graded
      quality
      A
      maintenance
      Enables MCP-compatible applications to interact with local and cloud-hosted LLM models for model management, text generation, chat, embeddings, and web search/fetch.
      817 npm
      AGPL 3.0