KO | EN

GitLite

Minimalist Open Source Repository Search Engine

gitlite — search

#codex #developer-tools #typescript #ai-agents #claude-code #python #mcp #react #cli #macos #ai-agent #deepseek-harness

MCP-Runware

★ 9 Open GitHub ↗

Model Context Protocol (MCP) server that provides lightning fast image and video generation tools using the Runware API.

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# Runware MCP Server

A powerful Model Context Protocol (MCP) server that provides lightning fast image and video generation tools using the [Runware API](https://runware.ai). This server supports both SSE (Server-Sent Events) transport for custom claude connector and direct claude desktop installation as well.

## Features

### **Image Generation Tools**
- **`imageInference`**: Full-featured image generation with advanced parameters
- **`photoMaker`**: Subject personalization with PhotoMaker technology
- **`imageUpscale`**: High-quality image resolution enhancement
- **`imageBackgroundRemoval`**: Background removal with multiple AI models
- **`imageCaption`**: AI-powered image description generation
- **`imageMasking`**: Automatic mask generation for faces, hands, and people

### **Video Generation Tools**
- **`videoInference`**: Text-to-video and image-to-video generation
- **`listVideoModels`**: Discover available video models
- **`getVideoModelInfo`**: Get detailed model specifications

### **Utility Tools**
- **`imageUpload`**: Upload local images to get Runware UUIDs
- **`modelSearch`**: Search and discover AI models on the platform

### **Smart Features**
- **Automatic Model Selection**: I2V uses `klingai:5@2`, T2V uses `google:3@1`
- **Input Validation**: Prevents Claude upload URL pasting and validates dimensions
- **Comprehensive Error Handling**: Clear error messages and guidance

## Demo

Watch the demo video to see the Runware MCP server in action:

https://github.com/user-attachments/assets/9732096b-8513-455c-9759-cc88363c42f9

## Architecture

- **Transport**: SSE (Server-Sent Events) for real-time communication
- **Framework**: [FastMCP](https://github.com/jlowin/fastmcp) with [Starlette](https://www.starlette.io/) web framework
- **Server**: [Uvicorn](https://www.uvicorn.org/) ASGI server
- **API**: Direct integration with [Runware's AI services](https://runware.ai)

## Prerequisites

- **Python**: 3.10 or higher
- **Runware API Key**: Get your API key from [Runware Dashboard](https://my.runware.ai)
- **Dependencies**: See `requirements.txt` or `pyproject.toml`

## Installation

### **1. Clone the Repository**
```bash
git clone https://github.com/Runware/MCP-Runware.git
cd MCP-Runware
```

### **2. Install Dependencies**
```bash
# Using uv (recommended)
uv venv
source .venv/bin/activate
uv pip install .

# Or using pip
pip install -r requirements.txt
```

### **3. Environment Setup**
Create a `.env` file in the project root:
```bash
RUNWARE_API_KEY=your_api_key_here

```

## Deployment Methods

### **Method 1: SSE Server (Recommended for Production)**

#### **Docker Deployment**
```bash
# Build the Docker image
docker build -t runware_mcp_sse .

# Run the container
docker run --rm -p 8081:8081 runware_mcp_sse
```

### **Method 2: MCP Install (Direct Integration)**

#### **Install in Claude Desktop**
```bash
# From the project directory
mcp install --with-editable . runware_mcp_server.py
```

## Model Recommendations

### **Image Generation**
- **Default**: `civitai:943001@1055701` (SDXL-based)
- **PhotoMaker**: `civitai:139562@344487` (RealVisXL V4.0)
- **Background Removal**: `runware:109@1` (RemBG 1.4)

### **Video Generation**
- **Image-to-Video (I2V)**: `klingai:5@2` (1920x1080)
- **Text-to-Video (T2V)**: `google:3@1` (1280x720)

You can find all additional models here: [Runware Models](https://my.runware.ai/models/all)

## Configuration

### **Environment Variables**
- `RUNWARE_API_KEY`: Your Runware API key (required)

### **Input Validation**
- Rejects Claude upload URLs (`https://files.*`). Claude tends to include base64 strings in its reasoning/thinking process, which rapidly fills the context window with garbage data. [Learn more about this issue](https://claude.ai/public/artifacts/0f28d79d-47bd-4fb8-bc25-e2699d78479f)
- Supports local file paths, public accessible URLs (make sure it has proper file extension such as JPG, PNG, WEBP, etc), and Runware UUIDs

## Support

- **Documentation**: [Runware API Docs](https://runware.ai/docs)
- **Models**: [Browse All Models](https://my.runware.ai/models/all)
- **Dashboard**: [Runware Dashboard](https://my.runware.ai)
- **Issues**: Create an issue in this repository
- **Email**: support@runware.ai

MCP-Runware

Explore Similar Repositories

Was this content helpful?

Recent Feedback

Download README