mcp-image-gen
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| QWEN_VAE | No | The Qwen-Image 2.1 VAE model file name in the ComfyUI install. | the Qwen-Image 2.1 file name |
| QWEN_CLIP | No | The Qwen-Image 2.1 CLIP model file name in the ComfyUI install. | the Qwen-Image 2.1 file name |
| QWEN_UNET | No | The Qwen-Image 2.1 UNet model file name in the ComfyUI install. | the Qwen-Image 2.1 file name |
| COMFYUI_URL | No | The URL of the ComfyUI server used by the local backend. | http://127.0.0.1:8188 |
| OPENAI_API_KEY | No | Optional OpenAI API key. When set, enables the cloud backend; without it, the default 'auto' backend goes directly to ComfyUI. | |
| IMAGE_OUTPUT_DIR | No | Directory where generated images are saved when the tool call does not specify an output path. | ~/Pictures/mcp-image-gen |
| OPENAI_IMAGE_MODEL | No | The OpenAI Images API model to use for the cloud backend. | gpt-image-1 |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| generate_imageA | Generates a PNG image and saves it to disk. Tries the OpenAI Images API first (when OPENAI_API_KEY is set); when it is rate-limited or unavailable, falls back automatically to Qwen-Image 2.1 on the local ComfyUI server. Returns the file paths and the backend used. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 1 tool
With a single tool there is no possibility of misselection or overlap. An agent's only choice is whether to call it at all, and the description makes the trigger conditions (generate a PNG from a prompt, with automatic backend fallback) unambiguous.
The single name follows a clean verb_noun snake_case convention (generate_image) that would scale predictably if more tools were added. No style mixing is possible with one tool.
One tool is thin even for a narrowly scoped server; the rubric treats 1-2 tools as borderline. It earns its place, but the surface feels underbuilt for a server whose apparent remit is image generation and management.
Generation is well covered, including a fallback backend, but there is no way to list, inspect, delete, or re-run previously generated images, and no way to check backend/API status. These are notable gaps an agent would hit when managing outputs, though core generation works.