Skip to main content
Glama
artokun

ComfyUI MCP

by artokun
OverviewSchemaSourceRelated ServersDiscussionsAdmin
TypeScript
Hybrid

Related Servers

Alternatives to ComfyUI MCP

Related Servers

  • comfyui-mcp

    sandyup
    A
    license
    A
    quality
    B
    maintenance
    MCP server + Claude Code plugin for ComfyUI — execute workflows, generate images, visualize pipelines, manage models, control VRAM, and explore custom nodes, all from your AI coding assistant.
    48
    34,360 npm
    1
    MIT

  • comfy-draftsman

    EnragedAntelope
    A
    license
    A
    quality
    A
    maintenance
    The MCP server that drafts ComfyUI workflows a human can actually read. It lets coding agents build, repair, port, validate, and run ComfyUI workflows against your own ComfyUI instance, delivering clean, organized, fully-labeled workflows.
    29
    2
    MIT

  • ComfyUI-APP-MCP

    Lotus0614
    A
    license
    Not graded
    quality
    A
    maintenance
    A custom node plugin for ComfyUI that encapsulates workflows as templates, enabling AI assistants (Claude, Cursor) to invoke ComfyUI for multimedia generation via the MCP protocol.
    25
    MIT

  • comfy-mcp

    cuzelac
    A
    license
    Not graded
    quality
    D
    maintenance
    MCP server that enables AI agents to control a local ComfyUI instance for image generation, allowing workflow understanding, parameter modification, execution, and model discovery.
    17 npm
    3
    Apache 2.0

  • comfyui-roundabout

    0c0
    A
    license
    Not graded
    quality
    B
    maintenance
    Exposes ComfyUI workflows as 12 MCP tools on ComfyUI's own port, letting AI agents generate, edit, and remove backgrounds from images plus create videos by passing semantic parameters instead of whole workflow graphs. Agents can list models, submit and track async tasks, fetch artifact URLs, and jump to a visual asset/task page.
    2
    MIT
Deploy Server
One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

TDQS

A3.8/5.0

Scored across 41 tools

Disambiguation3/5

Each tool has long, precise descriptions and almost every multi-mode tool uses action subcommands, so individual tools are rarely ambiguous. However several cross-tool boundaries blur: top-level get_workflow vs queue action:'get_workflow', multiple model download paths (download_model, comfy_cli models_download, apply_manifest), and multiple enqueue surfaces. An agent would need to read descriptions carefully to pick the right one, so overlap exists but can be resolved.

Naming Consistency3/5

The dominant pattern is snake_case verb_noun (get_workflow, save_workflow, generate_image, install_custom_node), which is good. But there are also bare-noun tools (queue, runpod, workspace, apps, batch, kitchen, bisect), a noun_verb outlier (runpod_watch), and an inconsistent train_* cluster (train_prepare_dataset, train_start, train_doctor). This is readable but not a single predictable convention.

Tool Count2/5

Forty-one top-level tools, each carrying multiple action subcommands, makes this a very heavy surface for an agent to navigate. While the domain is broad (ComfyUI, model training, RunPod, custom-node development), several tools could be consolidated (runpod/runpod_watch, queue/enqueue_workflow, list_packs/list_tools). It exceeds the reasonable 25+ threshold on the scale.

Completeness4/5

The server covers the ComfyUI lifecycle unusually well: workflow creation/validation/submission, queue/history management, model discovery/installation/deletion, custom-node management, training, RunPod control, and asset retrieval. Gaps are minor, such as no delete/rename for saved workflows and no direct cleanup of generated assets. Overall, agents can complete most user workflows without hitting dead ends.

Maintenance

ActivityActive
ResponsivenessResponsive

Resources