RunComfy

You run an MCP server that leverages RunComfy APIs to generate AI videos and images. This enables you to produce media assets programmatically through simple tool calls, integrating AI video and image generation into your workflow with security and repeatable configurations.

How to use

After you have the MCP server set up, you interact with it through your MCP client by calling the available tools. You can generate AI videos with descriptive prompts, choose from supported models, and adjust parameters like duration and aspect ratio. You can also generate AI images, edit existing images, check the status of a request, and retrieve results. Use these capabilities in sequences (for example, submit a video generation, poll for status, then fetch the final URL) to automate media creation workflows.

Key tools you can use include: generating a video from a prompt, generating an image from a prompt, editing an image, checking request status, retrieving results, canceling a queued request, and listing available models. Each tool accepts a set of parameters such as prompt, model, aspect_ratio, seed, and optional advanced inputs. Use the RunComfy API key you configure to authorize requests.

How to install

Prerequisites: You will need a working Node-like runtime (bun) installed on your system. Ensure you have a RunComfy API key ready for authentication.

cd mcp-servers/runcomfy
bun install

Obtain your API key from RunComfy and prepare to configure the MCP server with this key.

{
  "mcpServers": {
    "runcomfy": {
      "command": "bun",
      "args": [".../mcp-runcomfy/index.js"],
      "env": {
        "RUNCOMFY_API_KEY": "YOUR_API_KEY_HERE"
      }
    }
  }
}

Add the configuration to your Windsurf MCP config file at the path shown above, and then restart Windsurf to apply the changes.

# Restart Windsurf after saving the config
windsurf restart

Additional notes

The setup assumes you will store the MCP configuration in your user profile under a Windsurf configuration directory. It uses a standard structure where a named MCP server entry declares the runtime command, the arguments to run, and the environment variables needed for authentication.

Examples of common operations

// Generate video
runcomfy_generate_video({
  prompt: "A calm person breathing slowly, teal glow, dark background",
  model: "wan-2.1",
  aspect_ratio: "1:1"
})

// Check status
runcomfy_check_status({ request_id: "abc123" })

// Get result
runcomfy_get_result({ request_id: "abc123" })

// Generate image
runcomfy_generate_image({
  prompt: "A minimal flat illustration of a calm person breathing, teal and navy",
  model: "flux-2-pro",
  aspect_ratio: "1:1"
})

// Edit image
runcomfy_edit_image({
  prompt: "Change the background to a dark gradient and add a subtle teal glow",
  model: "flux-2-dev-edit",
  image_urls: ["https://example.com/input.png"]
})

Available tools

runcomfy_generate_video

Generate an AI video using RunComfy models with a specified prompt, model, and optional parameters like duration, aspect_ratio, and seed.

runcomfy_generate_image

Create an AI image from a text prompt using the flux-2-pro model, with optional aspect_ratio, seed, and advanced inputs.

runcomfy_edit_image

Edit an existing image by providing a prompt, a model suitable for image-to-image edits, and one or more image URLs.

runcomfy_check_status

Check the status of a submitted request by its request_id.

runcomfy_get_result

Retrieve the final result, such as a video URL, for a completed request.

runcomfy_cancel

Cancel a queued RunComfy request.

runcomfy_list_models

List available RunComfy models and curated model maps for video and image tasks.