fal

Fal.ai MCP Server enables Claude Desktop and other MCP clients to generate images, videos, music, and audio using Fal.ai models. It exposes a rich set of tools for media generation, editing, and utility tasks, including image generation (generate_image, generate_image_structured, generate_image_from_image), image editing (remove_background, upscale_image, edit_image, inpaint_image, resize_image, compose_images), video tools (generate_video, generate_video_from_image, generate_video_from_video), and audio tools (generate_music). It also offers model discovery and recommendations (list_models, recommend_model) along with usage tracking and file uploads (get_usage, get_pricing, upload_file). The server supports multiple transports (STDIO, HTTP/SSE) and can run in single or dual mode, making it suitable for local development or production deployments behind a reverse proxy or container runtime.

To use it, configure Claude Desktop (or any MCP client) with a mapped MCP server entry. For example, provide the command fal-mcp with an environment variable FAL_KEY containing your Fal.ai API key. Depending on your deployment, you may connect via STDIO for local development, or via the HTTP/SSE endpoint when running in a container or on a remote host. If you choose the Docker route, start the container with your API key and the desired transport, then point your MCP client at the exposed endpoint (e.g., http://localhost:8080/sse for SSE).

Prerequisites:\n- Python 3.10 or higher (recommended)\n- A Fal.ai API key (free tier available)\n- Optional: Claude Desktop or any MCP-compatible client\n\nInstallation options:\n1) Install from PyPI (recommended for local development)\nbash\npip install fal-mcp-server\n````\n2) Install with uv (zero-install alternative)\nbash\nuvx --from fal-mcp-server fal-mcp\n\n3) Install from source (advanced)\nbash\ngit clone https://github.com/raveenb/fal-mcp-server.git\ncd fal-mcp-server\npip install -e .\n\n4) Docker (production)\nbash\ndocker run -d --name fal-mcp -e FAL_KEY=your-api-key -p 8080:8080 ghcr.io/raveenb/fal-mcp-server:latest\n```\nAfter installation, obtain your Fal.ai API key from fal.ai and configure your MCP client as described in the configuration examples.

Environment variables and configuration tips:\n- FAL_KEY: Your Fal.ai API key must be provided to authenticate requests.\n- FAL_MCP_TRANSPORT, FAL_MCP_HOST, FAL_MCP_PORT (Docker): Control transport mode and binding.\n- When using Docker for file uploads, mount host directories into the container and reference them via container paths (e.g., /downloads, /pictures).\n- If connecting via HTTP/SSE, ensure port accessibility through firewalls or reverse proxies.\n- For Claude Desktop integration, the MCP server name in config is typically fal-ai; adjust as needed for your environment.\n- If you encounter permission or path issues with uploads, verify that the target paths are accessible by the MCP server process and that the required volume mounts are configured.