MCP 服务工具

What is MCP 服务工具?

MCP 服务工具 is a Model Context Protocol server that provides tool integrations for AI service APIs. It currently supports Hugging Face (text, image, speech, zero‑shot tasks) and Dify (conversational and text generation apps) services. It is aimed at developers who want to use MCP to access these AI models through a unified interface.

How to use MCP 服务工具?

Install dependencies with npm install, create a .env file with the required API keys (HUGGINGFACE_API_KEY, DIFY_API_KEY, DIFY_BASE_URL), then start the Dify MCP service using npm run start:dify. For Hugging Face, import and call the tool functions directly in TypeScript, providing the API key and model parameters.

Key features of MCP 服务工具

• Hugging Face: text generation, classification, QA, summarization, translation, fill‑mask
• Hugging Face: image classification, segmentation, image‑to‑text, zero‑shot image/text classification
• Hugging Face: automatic speech recognition
• Dify: chat with conversation history support
• Dify: text generation via completion endpoint
• Dify: conversation listing and renaming

Use cases of MCP 服务工具

• Build an MCP‑compliant chat interface that routes user messages to a Dify‑powered AI assistant.
• Perform on‑the‑fly text or image classification using Hugging Face models without managing separate endpoints.
• Integrate speech recognition into an MCP agent to transcribe audio blobs.
• Generate summaries or translations of long documents through MCP tool calls.
• Manage multiple Dify conversations (list, rename, fetch history) programmatically.

FAQ from MCP 服务工具

What are the runtime dependencies?

Node.js (npm) is required. The README does not specify a minimum version, but npm install must succeed.

What API keys are needed?

For Hugging Face, set HUGGINGFACE_API_KEY. For Dify, set DIFY_API_KEY and DIFY_BASE_URL. These are added in a .env file in the project root.

Are there any usage limits or costs?

Yes. The README notes that users should be aware of API call limits and associated costs for both Hugging Face and Dify. Exact limits depend on the provider’s plan.

Does the server support authentication?

Authentication is handled via API keys supplied in the environment. There is no mention of other auth methods (e.g., OAuth).

What transport does MCP 服务工具 use?

The README does not specify the MCP transport (e.g., stdio, SSE). It only shows a start command (npm run start:dify) and TypeScript API usage.