Replicate Flux MCP Server

How to use

With this MCP server, you can ask your AI assistant to create visuals from text prompts. It supports high-quality bitmap images using Flux Schnell and scalable SVG graphics using the Recraft V3 SVG model. Use it to generate single images, batches of variants, or multiple interpretations of the same concept. You can also request SVG vectors for scalable graphics workflows. To get started, ensure you have an active Replicate API token and a configured MCP client that can talk to your local stdio MCP endpoint.

How to install

Prerequisites you need before installing: node 18+ and npm, plus an active Replicate API token. You will run the MCP server locally and connect your MCP clients to it.

Step 1. Clone the project repository or obtain the MCP server package as instructed by your source distribution. Step 2. Install dependencies. Step 3. Start the MCP server in development or production mode as appropriate for your setup.

Configuration and usage notes

{
  "mcpServers": {
    "replicate_flux_cursor": {
      "command": "npx",
      "args": ["-y", "replicate-flux-mcp"],
      "env": {
        "REPLICATE_API_TOKEN": "YOUR_TOKEN"
      }
    },
    "replicate_flux_claude": {
      "command": "npx",
      "args": ["-y", "replicate-flux-mcp"],
      "env": {
        "REPLICATE_API_TOKEN": "YOUR_TOKEN"
      }
    }
  }
}

Troubleshooting

If authentication fails, double-check that your Replicate API token is correctly set in the environment for the MCP process and that the token is valid. If you encounter safety-related blocks from prompts, adjust the prompt to avoid restricted content. For timeouts on large generations, increase polling considerations in your runtime environment or adjust client side timeouts.

Examples

Here are practical prompts you can use with the server to generate images and SVGs. For multiple prompts in a single request, you can use batch generation to produce several outputs at once. For variations, you can request different interpretations of the same concept.

Technical details

This MCP server uses the Flux Schnell model for raster images and the Recraft V3 SVG model for vector graphics, interfaced via the Replicate API. It runs with a TypeScript/Node-based stack and validates inputs at runtime to ensure robust API interactions.

Examples

Batch Image Generation with generate_multiple_images: supply multiple prompts to receive several distinct images in one request.

Image Variants with generate_image_variants: request several variants of a single prompt to explore different takes on the same idea.

SVG generation with generate_svg: create vector graphics based on a descriptive prompt for scalable artwork.

Notes on deployment

This server is designed to run locally to maximize privacy of data and outputs. You can connect MCP clients like Cursor or Claude Desktop by pointing them at the configured stdio endpoints, using the provided start commands and environment variable requirements.

Developer guidance

If you plan to customize or extend the server, review the development workflow to clone the repository, install dependencies, run in development mode, and build for production. The typical flow includes installing dependencies, starting in dev mode, and building the project before deployment.

generate_image

Generates a single image from a text prompt using Flux Schnell with configurable seed, megapixels, aspect ratio, and quality.

generate_multiple_images

Generates multiple images from an array of prompts in one request, enabling batch generation.

generate_image_variants

Generates multiple variants of the same prompt, with optional prompt variations and varying seeds.

generate_svg

Generates an SVG vector image from a text prompt using the Recraft V3 SVG model.

prediction_list

Retrieves a list of recent Replicate predictions.

get_prediction

Fetches detailed information about a specific prediction by its ID.