Image MCP Servers

What is Image MCP Servers?

Image MCP Servers is an SSE-based MCP server that exposes tools for interacting with HuggingFace and Replicate APIs, focusing on image generation and model management. Built using the MCP Core framework, it provides a standardized interface for AI model interactions and is designed for developers who want to host their own integration layer.

How to use Image MCP Servers?

Clone the repository, copy .env.example to .env, add your HF_API_TOKEN and REPLICATE_API_TOKEN, then build with cargo build and run with cargo run. The server starts on the port specified by SERVER_PORT (default 3000). Optional Docker deployment is available via just docker-up service.

Key features of Image MCP Servers

• HuggingFace integration: get model info, sample images, README content, search models, and WhoAmI.
• Replicate integration: generate images, get model info, check prediction status, list models, and WhoAmI.
• Modular build with optional feature flags for HuggingFace or Replicate only.
• Configurable server name, version, port, and API tokens via environment variables.
• Docker support for containerized deployment.

Use cases of Image MCP Servers

• Build an AI assistant that can generate images on demand using Replicate.
• Search and retrieve model metadata from HuggingFace for curation or analysis.
• Create a unified interface for both HuggingFace and Replicate model management.
• Self-host a lightweight API gateway for image generation workflows.
• Integrate with MCP-compatible clients to expose model info and image generation tools.

FAQ from Image MCP Servers

What are the prerequisites to run Image MCP Servers?

Rust (2024 edition) is required. Docker is optional for containerized deployment.

How do I configure API tokens?

Create a .env file with HF_API_TOKEN="your_huggingface_token" and REPLICATE_API_TOKEN="your_replicate_token".

Can I use only one of the two services?

Yes. Build with --no-default-features --features huggingface or --no-default-features --features replicate to enable only one integration.

What port does the server run on?

The server runs on the port specified by the SERVER_PORT environment variable, defaulting to 3000.

Does the server require Docker?

No. Docker is optional; you can run the server natively with cargo run.