{
  "markdown": "# ComfyUI MCP Server for .NET\n\n[![Release](https://github.com/EntityAdam/comfyui-mcp/actions/workflows/release.yml/badge.svg)](https://github.com/EntityAdam/comfyui-mcp/actions/workflows/release.yml)\n[![NuGet](https://img.shields.io/nuget/vpre/ComfyUi.Mcp.svg)](https://www.nuget.org/packages/ComfyUi.Mcp)\n[![License: MIT](https://img.shields.io/github/license/EntityAdam/comfyui-mcp.svg)](LICENSE)\n\n<!-- mcp-name: io.github.EntityAdam/comfyui-mcp -->\n\nGive an AI agent access to ComfyUI generation, workflows, installed models,\nrecent assets, and job controls through the Model Context Protocol.\n\n## Quickstart\n\n### 1. Install .NET 10\n\nInstall the [.NET 10 SDK](https://dotnet.microsoft.com/download/dotnet/10.0).\nThe SDK includes `dnx`, which downloads and runs the pinned NuGet package\nwithout a permanent installation.\n\nStart ComfyUI normally and leave it available at `http://localhost:8188`.\n\n### 2. Paste the server configuration\n\nAdd this server to your MCP configuration:\n\n```json\n{\n  \"mcpServers\": {\n    \"comfyui\": {\n      \"command\": \"dnx\",\n      \"args\": [\n        \"ComfyUi.Mcp@0.0.1-preview5\",\n        \"--yes\",\n        \"--\",\n        \"--stdio\"\n      ]\n    }\n  }\n}\n```\n\nSee [MCP client configuration](docs/mcp-configuration.md) for a complete\nconfiguration and copy-paste alternatives, including connecting to ComfyUI at\nanother URL.\n\nSome clients call the top-level property `servers` instead of `mcpServers`.\nRestart or refresh the MCP client after saving the configuration.\n\n### 3. Let your agent finish setup\n\nAsk:\n\n> Run `setup_guide()` and help me configure ComfyUI MCP.\n\nThe guide discovers installed models and bundled workflow families, then helps\nthe agent choose workflows and create reusable defaults. The server can start\nwhile ComfyUI is offline, but model discovery and generation require ComfyUI to\nbe running.\n\n## Setup guide\n\nFor a more deliberate first setup, ask your agent to:\n\n1. Run `setup_guide()`.\n2. Run `workflow_list()` and show workflows that match your installed models.\n3. Run `family_list()` and `family_model_list()` to check model associations.\n4. Run `workflow_options_show()` before the first execution.\n5. Save any defaults you approve with `workflow_config_set()`.\n6. Generate a small test image and display it with `image_show()`.\n\nWorkflow JSON defines the available parameters. Optional YAML layers provide\nreusable family, workflow, and model defaults. Tool-call values always have the\nhighest precedence. See [Configuration](docs/configuration.md) for the complete\nmodel.\n\n## Features and usage\n\n### Families\n\nFamilies group compatible installed models and enable random selection.\n\n- `family_list()` discovers configured families.\n- `family_model_list()` shows installed members.\n- `family_model_add()` and `family_model_remove()` maintain exact membership.\n\nTry asking: “Show my configured families and which installed models belong to\neach one.”\n\n### Workflows\n\nBundled workflows cover image, audio, video, editing, upscaling, interpolation,\nand text generation.\n\n- `workflow_list()` discovers workflow IDs.\n- `workflow_options_show()` resolves inputs, defaults, and allowed values.\n- `workflow_execute()` runs any parameterized workflow.\n- `workflow_config_set()` saves reusable defaults.\n\nTry asking: “List the workflows I can run, then explain the options for the\nbest text-to-image workflow.”\n\n### Models\n\n`model_list()` reads the current checkpoint and diffusion-model inventory\ndirectly from ComfyUI. Exact model requests must use the returned filename,\nincluding its extension.\n\nTry asking: “List my installed models and recommend compatible workflow\nfamilies without inventing filenames.”\n\n### Assets\n\nGenerated outputs and their provenance are stored in a bounded SQLite registry\nthat survives MCP server restarts.\n\n- `asset_list()` and `asset_show()` inspect recent outputs.\n- `generation_show()` follows durable event-driven generation state.\n- `image_show()` renders an image in the conversation.\n- `asset_regenerate()` reuses a submitted workflow with new parameters.\n- `asset_publish_info()` and `asset_publish()` copy outputs into a project.\n\nTry asking: “Show my five most recent assets and display the newest image.”\n\n### ComfyUI\n\nThe server can upload inputs and manage ComfyUI jobs directly.\n\n- `comfy_media_upload()` stages image, audio, or video inputs.\n- `comfy_queue_status_show()` reports running and pending jobs.\n- `comfy_job_show()` reads local state for a prompt submitted by this server.\n- `comfy_job_cancel()` cancels queued or running work.\n\nTry asking: “Show the ComfyUI queue and explain what is currently running.”\n\n## How-to guides\n\n- [Generate audio from text](docs/how-to/text_to_audio.md)\n- [Generate an image from text](docs/how-to/text_to_image.md)\n- [Edit an image](docs/how-to/image_edit.md)\n- [Upscale an image](docs/how-to/image_upscale.md)\n- [Generate video from text](docs/how-to/text_to_video.md)\n- [Interpolate video frames](docs/how-to/video_frame_interpolation.md)\n\n## Tips and tricks\n\n### Pick a random model from a family\n\nAsk the agent to call `family_model_list()` first, then use\n`model_strategy: random` with the chosen family:\n\n```json\n{\n  \"prompt\": \"A cinematic city at dusk\",\n  \"workflow_id\": \"image_krea2_turbo_t2i\",\n  \"family\": \"krea2-turbo\",\n  \"model_strategy\": \"random\",\n  \"aspect_ratio\": \"16:9\"\n}\n```\n\nThe server selects only from installed models associated with that family.\n\n### Add `readme()` to your system prompt\n\nThe `readme()` tool returns compact operating instructions for agents, including\nworkflow discovery, exact model selection, media upload, and image display.\nCall it once and add the returned text to your agent's system prompt or\npersistent project instructions. `read_skill()` is an alias.\n\n## More documentation\n\n- [MCP client configuration](docs/mcp-configuration.md)\n- [Configuration](docs/configuration.md)\n- [Run from source](docs/run_from_source.md)\n- [Build and release](docs/build_and_release.md)\n\nSelf-contained Windows, Linux, and macOS archives are attached to each\n[GitHub release](https://github.com/EntityAdam/comfyui-mcp/releases).\n\n## Acknowledgements\n\nPorted technical requirements from\n[artokun/comfyui-mcp](https://github.com/artokun/comfyui-mcp). Thank you!\n",
  "bytes": 6217,
  "sha": "a9e8fbafeb4b0c458bffbed1739cf95fb9b43b0348ca95e5440a53ba92dfda63",
  "repo_slug": "entityadam/comfyui-mcp",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/mcp_io_github_entityadam_comfyui_mcp_9fee067e/readme"
}