Fal Image/Video

You can run a high-performance MCP server that connects FAL AI image and video generation to MCP-compatible clients. It automatically downloads generated content to your local machine, provides public and data URLs, and supports flexible delivery to clients like Claude Desktop. This guide walks you through using and installing the server, with practical steps to get you up and running quickly.

How to use

You interact with the server through MCP clients by starting either the local stdio server or the HTTP server mode. In stdio mode, Claude Desktop or compatible clients connect via standard input/output, while HTTP mode exposes endpoints over HTTP/S for remote clients. Your posts or prompts are sent to FAL AI via the MCP server, which then returns generated images or videos along with access URLs and download paths.

Key benefits you get from this MCP server include automatic local downloads of all generated content, three ways to access the results (public URLs, data URLs, and local file paths), and convenient environment options to customize download locations and how results are opened. This makes it easy to share results, embed them in other workflows, or view them immediately on your machine.

How to install

Prerequisites you need before installation are Node.js and npm. Ensure you have a modern Node.js runtime installed on your system.

1. Install or run the MCP server in stdio mode using NPX. This starts the server in the same way you would run other one-off MCP tools.

2. If you prefer HTTP mode for remote clients, you can start the HTTP server to expose MCP endpoints on a chosen port.

3. When using HTTP mode, set your FAL API key as an environment variable or pass it in the command line as shown in the configuration examples.

Configuration and usage notes

Two primary configuration profiles are provided for the MCP server. The stdio profile runs locally via npx fal-image-video-mcp and connects to clients through standard IO. The HTTP profile runs the server in HTTP mode, opening a REST-like interface and an MCP SSE endpoint for client communication.

Environment variables shown in examples control API access and download behavior. The key variables are FAL_KEY for authentication, DOWNLOAD_PATH for a custom download directory, ENABLE_DATA_URLS to enable or disable embedding as data URLs, MAX_DATA_URL_SIZE to cap data URL size, and AUTOOPEN to automatically open generated files with the default application.

The default download path is your user’s Downloads folder, but you can customize it through the DOWNLOAD_PATH variable when starting the server.

Security and access considerations

Treat your FAL API key as a secret. Only expose HTTP endpoints to trusted clients, and consider using network controls or authentication in production deployments. When using Lazy Authentication features, you can discover tools without an API key and authenticate when you invoke a specific tool.

Available tools

list_available_models

Discover all available models in the registry for image generation and related capabilities.

execute_custom_model

Run any FAL endpoint not in the curated registry by providing a custom endpoint and input parameters.