comfy-image-mcp
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| COMFYUI_URL | Yes | URL of the ComfyUI server, e.g. http://192.168.1.50:8188. Required. | |
| COMFY_TIMEOUT_MS | No | Timeout per ComfyUI job in milliseconds. Default: 600000. | |
| COMFY_MODEL_LABEL | No | Model label written to sidecars and tool output. Default: comfyui. | |
| COMFY_PROJECT_ROOT | No | Project root that outputs and refs must stay inside. Defaults to the first MCP file root reported by the client, else the directory the client started the server in. Must be set if the resolved root is /. | |
| COMFY_WORKFLOWS_DIR | No | Directory where txt2img.json, edit.json and cutout.json live. Defaults to the package's workflows/ folder. | |
| COMFY_MAX_MEGAPIXELS | No | Maximum requested image size in megapixels; larger requests are scaled down. Default: 2.1. | |
| COMFY_INLINE_MAX_BYTES | No | Largest PNG returned inline in the tool result, in bytes. 0 turns this off. Default: 3000000. |
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 | {
"listChanged": true
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| generate_imageA | Generates an image with a diffusion model running in ComfyUI on a GPU machine. No API key or per-image cost. Use it for design mockups, reference-guided variations and edits (pass |
| image_backend_statusA | Reports whether ComfyUI is reachable, GPU/VRAM, which workflows (txt2img, edit, cutout) are configured, and the project root images are written to. Does not generate anything. |
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 2 tools
The two tools are clearly separated: one is a read-only diagnostic (backend/GPU/workflow status) and the other performs generation. There is no plausible way to confuse them, and image_backend_status explicitly states it does not generate anything.
Both names are snake_case and readable, but they mix a verb_noun form (generate_image) with a noun_phrase form (image_backend_status). It's a minor stylistic deviation rather than a real inconsistency.
Two tools is thin even for a narrow ComfyUI wrapper; the status tool is essentially a prerequisite check for the single real action. It's defensible for the scope but sits at the borderline of under-provisioned.
Generation plus a health check covers the core happy path, including refs for edits and cutouts, but there is no way to list previously generated images, enumerate/select configured workflows, cancel long jobs, or post-process assets. Agents will hit dead ends when they need to inspect or manage outputs.