Skip to content
Back to search

ComfyUI-Agent-Kit

One ComfyUI skill for every AI coding agent (Claude Code, Codex, Gemini CLI, Qwen Code): drives a local ComfyUI end to end. 75 model prompt recipes, 581 templates, plus standalone Seedance, Krea and MiniMax H3 skills. Hardware-aware selection, workflow building, multi-shot video, ControlNet, HDR. By AI VFX NEWS.

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/slavasexton--comfyui-agent-kit/badge.svg)](/listings/slavasexton--comfyui-agent-kit)

HTML

<a href="/listings/slavasexton--comfyui-agent-kit"><img src="/api/listings/slavasexton--comfyui-agent-kit/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

Claude Code: one-command plugin

Claude Code users can add the kit straight from the marketplace, no clone needed:

/plugin marketplace add SlavaSexton/ComfyUI-Agent-Kit
/plugin install comfyui@comfyui-agent-kit

That registers the local `comfyui-mcp` driver (launched with `npx`, no manual npm step) and loads the full `comfyui` skill (the 75-entry recipe brain + the docs). You still need a local ComfyUI on `http://127.0.0.1:8188`; the skill fills in your machine block on the first task. Plugins are Claude Code only, so for **Codex / Gemini CLI / Qwen Code** use the multi-agent installer below.

Every agent: the installer

Windows (PowerShell):

git clone https://github.com/SlavaSexton/ComfyUI-Agent-Kit.git
cd ComfyUI-Agent-Kit
./install.ps1 -ComfyUIPath "E:\path\to\ComfyUI"   # installs for every agent CLI found on PATH

Linux / macOS:

git clone https://github.com/SlavaSexton/ComfyUI-Agent-Kit.git
cd ComfyUI-Agent-Kit
./install.sh --comfyui-path /path/to/ComfyUI       # installs for every agent CLI found on PATH

The installer runs the shared machine setup once (MCP package, templates, in-graph nodes), then **auto-detects** which of `claude` / `codex` / `gemini` / `qwen` are installed and wires each one. It is **idempotent**, re-run it any time. Limit the targets with `-Agents claude,gemini` / `--agents claude,gemini`. Flags: `-SkipTemplates` / `--skip-templates` (skip the ~900MB template clone), `-SkipNodes` / `--skip-nodes`. Per-agent details and the GLM note are in **[docs/AGENTS.md](docs/AGENTS.md)**.

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

.claude-plugin/marketplace.json · Manifest · 16 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

This file has changed since it was scanned, so this is not exactly the text that was scanned, and no line is marked. The findings are listed with the safety scan above.

  1. {
  2. "name": "comfyui-agent-kit",
  3. "description": "Local-first ComfyUI skills + the comfyui-mcp connection for Claude Code. The plugin form of ComfyUI-Agent-Kit, the local-first counterpart to the Comfy Cloud MCP.",
  4. "owner": {
  5. "name": "AI VFX NEWS",
  6. "url": "https://aivfxnews.com/"
  7. },
  8. "plugins": [
  9. {
  10. "name": "comfyui",
  11. "source": "./claude-code",
  12. "description": "Drive a LOCAL ComfyUI (your GPU, no cloud, no account) end to end: 75 model prompt recipes, hardware-aware model selection, workflow building, multi-shot video, the comfyui-mcp driver, plus a standalone Seedance skill for ByteDance video models. The local-first counterpart to the Comfy Cloud MCP.",
  13. "homepage": "https://github.com/SlavaSexton/ComfyUI-Agent-Kit"
  14. }
  15. ]
  16. }
README.md · README · 400 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

This file has changed since it was scanned, so this is not exactly the text that was scanned, and no line is marked. The findings are listed with the safety scan above.

  1. <div align="center">
  2. <img src="docs/assets/cover.png" width="880" alt="ComfyUI skill for AI coding agents, by Chronicle of Brothers">
  3. # ComfyUI-Agent-Kit
  4. **Local-first ComfyUI for every AI coding agent (Claude Code, Codex, Gemini CLI, Qwen Code).**
  5. <br>
  6. **Your GPU, your models, no cloud, no account.**
  7. **By [Chronicle of Brothers](https://chronicleofbrothers.com/).**
  8. Make Claude Code, Codex, Gemini CLI, or Qwen Code drive **ComfyUI** at full power on your own machine - generate
  9. images, video, and audio, build and run workflows, pick the model variant that fits *your* hardware, and **show the
  10. graph live in your own ComfyUI canvas**. No hosted service, no per-generation billing: one installer wires the same
  11. stack into every agent you run, then you hand the whole setup to someone else with one command.
  12. ![License: Apache 2.0](https://img.shields.io/badge/License-Apache_2.0-FFD27D.svg)
  13. ![ComfyUI](https://img.shields.io/badge/ComfyUI-driven-5BAEE3.svg)
  14. ![Agents](https://img.shields.io/badge/Claude_·_Codex_·_Gemini_·_Qwen-9aa3b2.svg)
  15. ![Platforms](https://img.shields.io/badge/Windows_·_Linux_·_macOS-supported-9aa3b2.svg)
  16. </div>
  17. ---
  18. This is the portable, machine-independent, **multi-agent** version of a working ComfyUI setup. One shared core
  19. (the knowledge + the MCP driver) plus a thin adapter per agent. Clone it, run the installer, and each of your
  20. agents gets the same stack, wired to *your* hardware. GLM (z.ai) run through Claude Code is covered by the
  21. `claude` adapter. See **[docs/AGENTS.md](docs/AGENTS.md)** for how each agent connects.
  22. **Local-first, for experts and everyday users alike.** It scales from one-command image / video generation to a
  23. professional VFX color pipeline: v2 ships **[ComfyUI-OCIO](https://github.com/SlavaSexton/ComfyUI-OCIO)** (nine
  24. Nuke-style OpenColorIO nodes - Read a sequence, grade in ACES, write ProRes, preview on an on-node player, fully color-managed) and the field
  25. guide to building your own custom nodes.
  26. > **Local-first by design.** Prefer the cloud? The official **Comfy Cloud MCP** runs your workflows on Comfy's
  27. > GPUs, no local setup. This kit is the local-first counterpart: everything runs on hardware *you* control, with no
  28. > account and no per-generation cost, the model picker sizes each job to your VRAM, and it serves four agents, not
  29. > one. Use whichever fits the job.
  30. ## What it can do
  31. - **Drive ComfyUI from four agents** (Claude Code, Codex, Gemini CLI, Qwen Code) off one shared core. GLM via
  32. Claude Code is covered too. ([docs/AGENTS.md](docs/AGENTS.md))
  33. - **~90-tool MCP driver.** The agent operates ComfyUI directly: generate, build / edit / validate graphs, queue,
  34. download models, manage VRAM, read logs, diagnose.
  35. - **Per-model "mega-brain":** 75 prompt-recipe entries covering 68 named models, distilled from **official
  36. sources** (image, video, audio, 3D), split by family so each file reads whole in one call;
  37. the agent auto-pulls the right recipe when you name a model, so it prompts each one in its own dialect.
  38. - **Knows where each model runs:** a [full index](docs/MODEL_INDEX.md) of all 163 library models (recipe /
  39. utility / template-only), local vs API.
  40. - **Hardware-aware model selection:** detects your VRAM, RAM, and free disk, then recommends the variant that
  41. fits (fp8 / offload / multi-GPU / quant) and refuses a download that won't fit, before wasting the bandwidth.
  42. - **18 enhancement and utility tools:** upscale / restore (Real-ESRGAN, SUPIR, SeedVR2), frame interpolation
  43. (FILM, RIFE), segmentation / depth / pose (SAM3, BiRefNet, Depth Anything), plus restoration chains.
  44. - **581-template library** (and **94 official Subgraph Blueprints**, reusable subgraph bricks) as the source of truth, plus **fetch any shared workflow by hash** and a **model
  45. shootout** (run a prompt through many models small, pick the winner, then scale up).
  46. - **Assembles new workflows from parts:** decomposes a task into stages, mixes templates and blueprint subgraphs,
  47. and wires the nodes correctly (output-to-input by type, with converters where needed), validated against
  48. `/object_info` before running. Not a preset runner.
  49. - **Expert color + custom nodes (new in v2):** ships **[ComfyUI-OCIO](https://github.com/SlavaSexton/ComfyUI-OCIO)**
  50. - nine Nuke-style OpenColorIO nodes (Read / Write a still, sequence or video, grade in ACES, write ProRes / EXR,
  51. preview on an on-node player, fully color-managed) - and the agent knows each node's I/O plus the field guide to *building* a custom node pack.
  52. ([docs/NODE_LIBRARY/ocio.md](docs/NODE_LIBRARY/ocio.md), [docs/BUILDING_NODES.md](docs/BUILDING_NODES.md))
  53. - **Starts ComfyUI for you:** when the server is down, the agent launches it headless in the background and
  54. generates (no need to open the app first); to peek, you open `http://127.0.0.1:8188` in a browser. For an
  55. unattended pipeline the start policy is configurable per project (env vars or a `.comfyui-agent.json`), so it
  56. never blocks on a prompt.
  57. - **GUI bridge + persistence:** the agent writes graphs into your ComfyUI canvas, and SAVES every workflow it
  58. builds or runs to ComfyUI's workflows folder, so you can open it later from the Workflows sidebar (an API
  59. generation alone leaves no trace on the canvas).
  60. - **Stays current on its own:** `check_updates.py` diffs the template repo and reads the blog RSS; an optional
  61. weekly task adds recipes for new models and pushes them. ([docs/UPDATING.md](docs/UPDATING.md))
  62. - **Three model skills that work outside ComfyUI too: `seedance`, `minimax-h3`, `krea`.** Vendor prompting
  63. knowledge that also applies on the vendor's own app or API ships as its own skill rather than being buried in
  64. the ComfyUI reference. `minimax-h3` owns the H3 three-field prompt format, `<d>` dialogue and the quant and
  65. acceleration ladder; `krea` owns the fork between Krea's hosted API and its open weights, FLUX.1 Krea Dev, and
  66. why the only ComfyUI pack for Krea Realtime is a 0-star lead rather than a route. The first of them, `seedance`: ByteDance's Seedance video models
  67. run on Dreamina, Jimeng and the BytePlus API as well as in ComfyUI, so their prompting knowledge ships as its
  68. own skill: the three task types and the one word that switches between them, the reference label syntax, the
  69. timestamp storyboard, the real-person character formula, and a failure table for face drift, duplicate
  70. characters and extension seams. Distilled from ByteDance's own guides.
  71. ([shared/seedance/SKILL.md](shared/seedance/SKILL.md))
  72. - **Portable and idempotent:** one installer, auto-detects your agents, re-runnable. Apache-2.0, no vendored third-party
  73. code (everything heavy is fetched at install).
  74. ## The four-layer stack
  75. <div align="center">
  76. <img src="docs/assets/architecture.png" width="880" alt="The four-layer stack: knowledge + client, MCP driver, in-graph Claude nodes, node-building skills, plus the template library and GUI bridge">
  77. </div>
  78. <br>
  79. | Layer | What | Installed as |
  80. |------:|------|--------------|
  81. | 1 | **Knowledge + client** the operating manual and a zero-dependency HTTP client | the agent's skill / extension dir |
  82. | 2 | **MCP driver** ~90 structured tools so the agent operates ComfyUI directly | `comfyui-mcp` (npm) + per-agent MCP registration |
  83. | 3 | **In-graph Claude nodes** an LLM as a step inside a workflow (prompt enrichment, vision QA) | ComfyUI `custom_nodes` |
  84. | 4 | **Node-building skills** for writing/modifying custom nodes (V3 API) | the agent's skill dir (Claude/Codex) |
  85. | + | **Template library** the official 500+ workflow templates, the source of truth | sparse git clone + quick index |
  86. Plus a **GUI bridge**: the agent writes graphs to `<ComfyUI>/user/default/workflows/`, you open them in the
  87. built-in Workflows sidebar and tweak them. No extra "agent panel" node required.
  88. See [docs/LAYERS.md](docs/LAYERS.md) for each layer, and [docs/AGENTS.md](docs/AGENTS.md) for the per-agent matrix.
  89. ## The template library is the source of truth
  90. The kit clones the official [Comfy-Org/workflow_templates](https://github.com/Comfy-Org/workflow_templates) and
  91. builds a compact lookup index so the agent can match any request to the right template. 581 templates (plus **94 official Subgraph Blueprints**, reusable subgraph bricks) span every
  92. task, image, video, 3D, audio, utilities:
  93. <div align="center">
  94. <img src="docs/assets/templates_by_category.png" width="880" alt="Workflow templates by category: 160 video, 158 image, 103 use cases, 72 utility, 33 3D, 31 audio, 17 LLM, 7 node basics">
  95. </div>
  96. ## It knows every model's dialect
  97. Each generative model rewards a different prompt approach: SDXL wants comma tags, FLUX wants natural-language
  98. sentences, video models want camera and motion direction, audio models want genre/tempo/instruments, and
  99. negative-prompt support varies wildly. The kit ships **[`MODELS.md`](shared/comfyui/MODELS.md)**, a per-model
  100. prompting reference distilled from **official sources** (each maker's docs and model cards, docs.comfy.org, and
  101. the per-model templates from the `anthropic-claude` node). When you name a model in a request or a workflow,
  102. the agent reads that model's entry first and prompts it correctly.
  103. Covered today (75 recipe entries, 68 named models): FLUX.1/.2 + Kontext, Z-Image, Boogu, Mage-Flow, Qwen-Image/Edit, SDXL, SD1.5/3.5, HiDream,
  104. Ideogram, Nano Banana Pro/2, Seedream, Recraft, GPT-Image, Grok, Reve, Kandinsky, BRIA, OmniGen, Chroma, Krea 1/2,
  105. ERNIE-Image, FireRed/LongCat/ChronoEdit/JoyAI Image Edit (edit), Capybara, Bernini-R, Anima, NewBie, PixelDiT, Ovis-Image, Lens,
  106. Quiver, Wan 2.1-2.7, LTX-2.3/2 Pro, Hunyuan Video, SVD, Kling, Veo, Sora, Seedance, Luma, Runway, MiniMax, PixVerse,
  107. Vidu, Pika, HeyGen (avatar video), HappyHorse, HuMo, SCAIL-2, Stable Audio, ACE-Step, ElevenLabs, ChatterBox, Sonilo, Hunyuan3D, Tripo,
  108. Rodin, Meshy. Plus a separate **Enhancement and utility** section (not prompt-driven, settings not prompts):
  109. upscalers and restorers (Real-ESRGAN, SUPIR, SeedVR2, FlashVSR, Topaz, Magnific), frame interpolation (FILM, RIFE),
  110. conditioning helpers (SAM3, BiRefNet, Depth Anything, DWPose, MoGe, IP-Adapter, LivePortrait, Mediapipe), and video
  111. object removal (VOID). Anything else falls back to the template library.
  112. <div align="center">
  113. <img src="docs/assets/models_by_modality.png" width="880" alt="Per-model prompt recipes by modality: 43 image, 25 video, 6 audio, 4 3D, 78 total, split local/open-weight vs API, plus 18 enhancement and utility tools">
  114. </div>
  115. **Full model index**: every model in the library and exactly what the kit has for it (recipe / utility /
  116. template-only): **[docs/MODEL_INDEX.md](docs/MODEL_INDEX.md)**.
  117. ### Coverage table: every model and whether a prompt recipe is ready
  118. `✅ recipe` = a dedicated, up-to-date prompting guide reached through the index in
  119. [MODELS.md](shared/comfyui/MODELS.md), which names the family file under `shared/comfyui/MODELS/` that holds
  120. it. `🔧 tool` = an enhancement/utility note (settings, not prompts). **Table last reviewed: 2026-08-06.**
  121. One table, columns aligned to the widest row (the video models).
  122. | Modality | Model / tool | Prompt recipe | Runs |
  123. |---|---|:---:|---|
  124. | Image | FLUX.1 / FLUX.2 / Kontext | ✅ | local + API |
  125. | Image | Z-Image-Turbo | ✅ | local |
  126. | Image | Qwen-Image / Edit | ✅ | local |
  127. | Image | SDXL · SD 1.5 · SD 3.5 | ✅ | local |
  128. | Image | HiDream-I1 | ✅ | local |
  129. | Image | BRIA 3.x | ✅ | local |
  130. | Image | OmniGen v1/v2 | ✅ | local |
  131. | Image | Chroma | ✅ | local |
  132. | Image | Krea 2 / FLUX.1 Krea Dev | ✅ | local |
  133. | Image | ERNIE-Image | ✅ | local |
  134. | Image | Capybara (image+video) | ✅ | local |
  135. | Image | Bernini-R (relight) | ✅ | local |
  136. | Image | Anima (anime, + ControlNet-LLLite) | ✅ | local |
  137. | Image | NewBie (anime, XML prompts) | ✅ | local |
  138. | Image | PixelDiT | ✅ | local |
  139. | Image | Ovis-Image (text rendering) | ✅ | local |
  140. | Image | Lens / Lens Turbo | ✅ | local |
  141. | Image | Quiver (text to SVG) | ✅ | API |
  142. | Image | Ideogram 2/3 | ✅ | API |
  143. | Image | Nano Banana Pro / 2 | ✅ | API |
  144. | Image | Seedream 4/5 | ✅ | API |
  145. | Image | Recraft V3 | ✅ | API |
  146. | Image | GPT-Image | ✅ | API |
  147. | Image | Grok Image | ✅ | API |
  148. | Image | Reve | ✅ | API |
  149. | Image | Kandinsky 3.x | ✅ | local + API |
  150. | Image edit | FireRed / LongCat / ChronoEdit | ✅ | local |
  151. | Image edit | JoyAI Image Edit (JD) | ✅ | local |
  152. | Video | Wan 2.1-2.7 (+VACE/Animate v1/Animate 2/ATI) | ✅ | local + API |
  153. | Video | LTX-2.3 / LTX-2 Pro | ✅ | local |
  154. | Video | Hunyuan Video | ✅ | local |
  155. | Video | SVD (image-to-video) | ✅ | local |
  156. | Video | HuMo (lip-sync) | ✅ | local |
  157. | Video | SCAIL-2 (character) | ✅ | local |
  158. | Video | HappyHorse 1.1 (synced audio) | ✅ | API |
  159. | Video | Kling (1.6-3.0, O1/O3) | ✅ | API |
  160. | Video | Veo 3/3.1 | ✅ | API |
  161. | Video | Sora 2 | ✅ | API |
  162. | Video | Seedance 1.0/1.5/2.0 (4K) | ✅ | API |
  163. | Video | Luma Ray · Runway Gen-4/4.5 | ✅ | API |
  164. | Video | MiniMax/Hailuo · PixVerse · Vidu · Pika | ✅ | API |
  165. | Video | HeyGen (avatar, talking photo, translate) | ✅ | API |
  166. | Audio | Stable Audio · ACE-Step · ChatterBox | ✅ | local |
  167. | Audio | ElevenLabs · Sonilo | ✅ | API |
  168. | 3D | Hunyuan3D | ✅ | local |
  169. | 3D | Tripo · Rodin · Meshy | ✅ | API |
  170. | Enhance / utility | Real-ESRGAN, SUPIR, SeedVR2, FlashVSR (upscale/restore) | 🔧 settings | local |
  171. | Enhance / utility | Topaz, Magnific (upscale) | 🔧 settings | API |
  172. | Enhance / utility | FILM, RIFE (frame interpolation) | 🔧 settings | local |
  173. | Enhance / utility | SAM3, BiRefNet (segmentation/matting) | 🔧 settings | local |
  174. | Enhance / utility | Depth Anything v2/v3, MoGe (depth/geometry) | 🔧 settings | local |
  175. | Enhance / utility | DWPose, Mediapipe (pose/landmarks) | 🔧 settings | local |
  176. | Enhance / utility | IP-Adapter, LivePortrait (conditioning/portrait) | 🔧 settings | local |
  177. | Enhance / utility | VOID (video object removal) | 🔧 settings | local |
  178. Niche models still without a recipe (very new, thin docs) run from their template and borrow the closest family's
  179. approach; see [docs/MODEL_INDEX.md](docs/MODEL_INDEX.md) for the full per-variant breakdown.
  180. ## Prerequisites
  181. - One or more agent CLIs on PATH: [Claude Code](https://claude.com/claude-code) (`claude`),
  182. [Codex](https://developers.openai.com/codex) (`codex`), [Gemini CLI](https://github.com/google-gemini/gemini-cli)
  183. (`gemini`), [Qwen Code](https://github.com/QwenLM/qwen-code) (`qwen`)
  184. - [Node.js](https://nodejs.org) (`node` + `npm`)
  185. - [git](https://git-scm.com), [Python 3](https://python.org)
  186. - A local **ComfyUI** install (Desktop or source), [comfy.org](https://www.comfy.org/)
  187. ## Install
  188. ### Claude Code: one-command plugin
  189. Claude Code users can add the kit straight from the marketplace, no clone needed:
  190. ```
  191. /plugin marketplace add SlavaSexton/ComfyUI-Agent-Kit
  192. /plugin install comfyui@comfyui-agent-kit
  193. ```
  194. That registers the local `comfyui-mcp` driver (launched with `npx`, no manual npm step) and loads the full
  195. `comfyui` skill (the 75-entry recipe brain + the docs). You still need a local ComfyUI on `http://127.0.0.1:8188`; the
  196. skill fills in your machine block on the first task. Plugins are Claude Code only, so for **Codex / Gemini CLI /
  197. Qwen Code** use the multi-agent installer below.
  198. ### Every agent: the installer
  199. Windows (PowerShell):
  200. ```powershell
  201. git clone https://github.com/SlavaSexton/ComfyUI-Agent-Kit.git
  202. cd ComfyUI-Agent-Kit
  203. ./install.ps1 -ComfyUIPath "E:\path\to\ComfyUI" # installs for every agent CLI found on PATH
  204. ```
  205. Linux / macOS:
  206. ```bash
  207. git clone https://github.com/SlavaSexton/ComfyUI-Agent-Kit.git
  208. cd ComfyUI-Agent-Kit
  209. ./install.sh --comfyui-path /path/to/ComfyUI # installs for every agent CLI found on PATH
  210. ```
  211. The installer runs the shared machine setup once (MCP package, templates, in-graph nodes), then **auto-detects**
  212. which of `claude` / `codex` / `gemini` / `qwen` are installed and wires each one. It is **idempotent**, re-run it
  213. any time. Limit the targets with `-Agents claude,gemini` / `--agents claude,gemini`. Flags: `-SkipTemplates` /
  214. `--skip-templates` (skip the ~900MB template clone), `-SkipNodes` / `--skip-nodes`. Per-agent details and the GLM
  215. note are in **[docs/AGENTS.md](docs/AGENTS.md)**.
  216. ## First run on a new machine
  217. After install, start ComfyUI, then in an agent session tell it to run the **bootstrap** once
  218. ([docs/BOOTSTRAP.md](docs/BOOTSTRAP.md)): it detects your GPUs, VRAM, RAM, free disk, paths, and installed models
  219. via the MCP `health_check`, fills the machine-specific block in the skill, and does a smoke-test generation. After
  220. that, just ask for media. On Claude/Codex the skill auto-activates on ComfyUI keywords; on Gemini/Qwen the
  221. knowledge is loaded as the extension's context.
  222. ## Optional: in-graph LLM key
  223. Only needed if you want a workflow to enrich prompts **without** the agent in the loop (e.g. an unattended pipeline):
  224. ```powershell
  225. setx CLAUDE_API_KEY "sk-ant-..." # then restart ComfyUI
  226. ```
  227. See [docs/NODES.md](docs/NODES.md). When you are driving, the agent writes prompts directly, better and free.
  228. ## Layout
  229. ```
  230. ComfyUI-Agent-Kit/
  231. ├── install.ps1 / install.sh top-level: shared setup + auto-detect agents + run adapters
  232. ├── shared/
  233. │ ├── comfyui/ SKILL.md + MODELS.md + comfy_client.py (one source of truth)
  234. │ └── tools/gen_quick_index.py rebuild the template lookup index
  235. ├── agents/
  236. │ ├── claude/ install.ps1/.sh -> ~/.claude/skills/comfyui + claude mcp add + CLAUDE.md
  237. │ ├── codex/ install.ps1/.sh -> ~/.agents/skills/comfyui + ~/.codex/config.toml
  238. │ ├── gemini/ install.ps1/.sh -> ~/.gemini/extensions/comfyui (gemini-extension.json + GEMINI.md)
  239. │ └── qwen/ install.ps1/.sh -> ~/.qwen/extensions/comfyui (qwen-extension.json + QWEN.md)
  240. ├── docs/AGENTS.md per-agent matrix (how each connects) + GLM note
  241. ├── docs/MODEL_INDEX.md every model in the library and what the kit has for it
  242. ├── docs/EXAMPLE_WORKFLOWS.md notable shared workflows (model shootouts, restoration) + fetch helper
  243. ├── docs/UPDATING.md stay current: check_updates.py (templates diff + blog RSS) + the loop
  244. ├── docs/BOOTSTRAP.md / LAYERS.md / NODES.md
  245. ├── ATTRIBUTION.md credits for fetched third-party pieces
  246. ├── CHANGELOG.md curated history of notable changes (Keep a Changelog)
  247. ├── CONTRIBUTING.md what the kit accepts, the Teaching standard, house rules
  248. ├── LICENSE Apache-2.0 (this kit's original files)
  249. └── NOTICE attribution + trademark notice (Apache-2.0 section 4d)
  250. ```
  251. ## What is and isn't in this repo
  252. In the repo (original work, Apache-2.0): the skill, the client, the installer, the index generator, the docs, the
  253. generated visuals. Fetched at install time from their own sources (not redistributed here): the `comfyui-mcp`
  254. package, the node-building skills, the workflow templates, and the in-graph Claude nodes.
  255. ## Credits and thanks
  256. **Our own companion pack:** **[ComfyUI-OCIO](https://github.com/SlavaSexton/ComfyUI-OCIO)** - nine Nuke-style
  257. OpenColorIO nodes - is by **Slava Sexton**, this kit's author (Apache-2.0). It is credited to him wherever the kit uses,
  258. recommends, or builds on it; see **[ATTRIBUTION.md](ATTRIBUTION.md)**.
  259. This kit stands on excellent open-source work. It is a thin wiring layer over these projects, and the heavy
  260. lifting is theirs. Huge thanks to:
  261. - **[ComfyUI](https://github.com/comfyanonymous/ComfyUI)** by comfyanonymous / Comfy-Org, the engine everything
  262. runs on.
  263. - **[comfyui-mcp](https://github.com/artokun/comfyui-mcp)** by [artokun](https://github.com/artokun), the MCP
  264. driver (Layer 2) that lets the agent operate ComfyUI with structured tools.
  265. - **[comfyui-custom-node-skills](https://github.com/jtydhr88/comfyui-custom-node-skills)** by
  266. [jtydhr88 / Terry Jia](https://github.com/jtydhr88), the node-building skills (Layer 4).
  267. - **[workflow_templates](https://github.com/Comfy-Org/workflow_templates)** by Comfy-Org, the template library
  268. that is the source of truth.
  269. - **[comfy-skills](https://github.com/Comfy-Org/comfy-skills)** by Comfy-Org, whose output-node validation guard
  270. and multi-reference compositing technique this kit adapts (for the local stack) in SKILL.md and ADVANCED.md.
  271. - **[comfyui-anthropic-claude](https://github.com/alexmunteanu/comfyui-anthropic-claude)** by
  272. [alexmunteanu](https://github.com/alexmunteanu) and
  273. **[comfyui_claude_prompt_generator](https://github.com/PauldeLavallaz/comfyui_claude_prompt_generator)** by
  274. [PauldeLavallaz](https://github.com/PauldeLavallaz), the in-graph Claude nodes (Layer 3).
  275. v1.1.0 builds on more excellent work. Thanks also to:
  276. - **[Prompt Relay](https://github.com/GordonChen19/Prompt-Relay)** by Gordon Chen, Ziqi Huang, and Ziwei Liu
  277. (S-Lab, NTU), the training-free temporal prompt-routing method (arXiv 2604.10030).
  278. - **[ComfyUI-PromptRelay](https://github.com/kijai/ComfyUI-PromptRelay)** and
  279. **[ComfyUI-SUPIR](https://github.com/kijai/ComfyUI-SUPIR)** by [kijai](https://github.com/kijai), the ComfyUI
  280. ports this kit recommends and drives.
  281. - **[LTX Director 2.0](https://github.com/WhatDreamsCost/WhatDreamsCost-ComfyUI)** by WhatDreamsCost, the LTX-2.3
  282. timeline-editor node.
  283. - **[Z-Image-Turbo Fun-ControlNet-Union](https://huggingface.co/alibaba-pai/Z-Image-Turbo-Fun-Controlnet-Union-2.1)**
  284. by alibaba-pai (PAI), plus the **LTX-2.3** model and **HDR IC-LoRA** by [Lightricks](https://huggingface.co/Lightricks).
  285. - **[Real-ESRGAN](https://github.com/xinntao/Real-ESRGAN)** by Xintao Wang and the BasicSR team, and **SUPIR** by
  286. the XPixel Group (Fanghua Yu et al.), the upscale and restore models. Note: the SUPIR weights are non-commercial.
  287. Field techniques in wide community use lean on:
  288. - **[KJNodes](https://github.com/kijai/ComfyUI-KJNodes)** by [kijai](https://github.com/kijai) (LTX-2.3 NAG, GGUF
  289. loading, chunked feed-forward, multi-guide), **[ComfyUI-CacheDiT](https://github.com/Jasonzzt/ComfyUI-CacheDiT)** by
  290. Jasonzzt (inference caching), **ComfyUI-MelBandRoFormer** (audio stem separation),
  291. **[ComfyUI-Frame-Interpolation](https://github.com/Fannovel16/ComfyUI-Frame-Interpolation)** by Fannovel16 (FILM),
  292. **comfyui-inpaint-cropandstitch** (Flux.2 masked inpaint), and
  293. **[GAP LTX 2.3 Motion](https://github.com/GeekatplayStudio/LTX-2-3-LipSync)** by GeekatplayStudio
  294. (lipsync / storyboard / long audio).
  295. - **[ComfyUI-Flux2Klein-Enhancer](https://github.com/capitan01R/ComfyUI-Flux2Klein-Enhancer)** by capitan01R, the
  296. training-free multi-reference identity-transfer node suite for FLUX.2 Klein. Note: PolyForm Noncommercial license
  297. (commercial use needs a separate license).
  298. - **[Smart Image Crop and Stitch](https://github.com/HallettVisual/ComfyUI-Smart-Image-Crop-and-Stitch)** by
  299. HallettVisual, an auto-sized crop/stitch node pair for high-res inpainting and detail edits (Apache-2.0).
  300. v2.5.0 leans on more people's work. Thanks also to:
  301. - **[Anima ControlNet-LLLite](https://huggingface.co/kohya-ss/Anima-LLLite)** by
  302. [kohya-ss](https://huggingface.co/kohya-ss), the control and inpainting patches that give the Anima base model
  303. depth, lineart, pose, scribble and masked editing (repacked for ComfyUI by Comfy-Org). Note: non-commercial
  304. license, inherited from the Anima base weights.
  305. - **[krea2_style_reference](https://huggingface.co/ostris/krea2_turbo_style_reference)** and
  306. **[ComfyUI-Krea2-Ostris-Edit](https://github.com/ostris/ComfyUI-Krea2-Ostris-Edit)** by
  307. [ostris](https://github.com/ostris), the Krea 2 Turbo image-style-reference LoRA and the instruction-edit nodes
  308. this kit documents, plus the **krea2-detail-enhancer-edit-lora** by
  309. [reverentelusarca](https://huggingface.co/reverentelusarca).
  310. - **[JoyAI-Image](https://github.com/jd-opensource/JoyAI-Image)** by JD (jd-opensource), the instruction image-edit
  311. model now supported natively in ComfyUI core (Apache-2.0).
  312. Full per-component licensing is in [ATTRIBUTION.md](ATTRIBUTION.md). If anything here misattributes your work,
  313. open an issue and it will be fixed.
  314. ## Contributing
  315. Read [CONTRIBUTING.md](CONTRIBUTING.md) first. It says what the kit accepts, what it does not and why, and the
  316. bar any knowledge entry has to clear. An issue before a large PR is welcome and usually faster.
  317. ## License
  318. Apache-2.0, see [LICENSE](LICENSE) and [NOTICE](NOTICE). Releases up to v3.1.0 were MIT and that grant stands. Third-party components keep their own licenses.
  319. <div align="center">
  320. Made by **[Chronicle of Brothers](https://chronicleofbrothers.com/)**
  321. </div>

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.