Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add instructions/germaniu/comfyui-mcp/agents-mdgit clone --depth 1 https://github.com/GermaniU/comfyui-mcpWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/instructions/germaniu/comfyui-mcp/agents-md)<a href="https://agentmods.dev/instructions/germaniu/comfyui-mcp/agents-md"><img src="https://agentmods.dev/badge/instructions/germaniu/comfyui-mcp/agents-md.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00900 | $0.00900 |
| Opus 5 | $0.00450 | $0.00450 |
| Sonnet 5 | $0.00180 | $0.00180 |
| Haiku 4.5 | $0.00090 | $0.00090 |
Grade A, and why
comfyui-mcp AGENTS.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 4d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 70 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md — comfyui-mcp
MCP HTTP/SSE server que wrappea ComfyUI (SDXL, RTX 3060) para generación de imágenes. Corre en Pop!_OS junto con ComfyUI (puerto 8188) y expone tools vía HTTP en el puerto 8201. Cualquier gateway Hermes de la LAN lo consume sin instalar nada localmente.
Stack
- Python 3.11+,
fastmcp,httpx,uvicorn,starlette - ComfyUI API en
http://127.0.0.1:8188(loopback, misma máquina) - GPU Broker (
~/stack/gpu-broker/gpu-broker.sh) coordina la GPU con llama-server
Estructura (vertical slice + clean)
src/comfyui_mcp/
├── config.py # env vars, presets, aspects, constantes
├── comfy_client.py # HTTP client thin a ComfyUI (system_stats, prompt, history, view)
├── gpu_arbiter.py # ensure_comfyui_running + wake
├── workflow.py # build_workflow (txt2img + img2img + upscale)
├── tools/ # generate_image, list_models, comfy_health, comfy_view_url, history
│ ├── __init__.py
│ ├── generate.py
│ ├── models.py
│ ├── health.py
│ └── view.py
└── server.py # FastMCP + entry point (stdio/HTTP)
tests/ # pytest por módulo
Reglas de estilo (heredadas del AGENTS.md de ComfyUI)
- Cambios chicos y directos. Tocar el camino de código más angosto que explica el problema.
- Cambiar la menor cantidad de archivos posible.
- Preferir fixes prácticos sobre arquitectura amplia. Abstracciones solo cuando quitan lógica repetida real.
- Preferir menos dependencias. No agregar deps a menos que sean necesarias.
- Borrar código obsoleto agresivamente. Sin ramas muertas, sin funciones nunca llamadas.
- Preservar APIs existentes, nombres de nodos, layout de archivos y compatibilidad de workflows.
- El código debe verse hand-written. Cambios que lean como código genérico de IA serán rechazados: capas helper innecesarias, nombres vagos, comentarios boilerplate, ramas defensivas sin failure mode real, rewrites amplios.
Tools actuales
| Tool | Descripción |
|---|---|
generate_image |
txt2img con presets (producto/realista/rapido/anime) o params manuales |
list_models |
Checkpoints, loras y samplers disponibles |
comfy_health |
Estado de ComfyUI: versión, VRAM, cola |
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 4d ago First seen · 70 lines · 900 tokens per session scan A dba7c9a341ae
comfyui-mcp AGENTS.md is an instructions file published in the GitHub repository GermaniU/comfyui-mcp (0 stars, last pushed 2d ago), licensed MIT. It adds 900 tokens to every session, about $0.0045 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
linkedin-content-engine CLAUDE.md
Claude Code instructions for mdbailin/linkedin-content-engine, covering claude implementation brief, goal, architecture constraints, required mcp tools and readbrandprofile.
docvideoer CLAUDE.md
Instructions for coding-ax/docvideoer, covering claude.md — docvideoer, 触发方式, 外部 skill 依赖, 工作目录 and 核心流水线（6 步）.
ai-workflow-kit CLAUDE.md
Claude Code instructions for bezael/ai-workflow-kit, covering claude.md — ai workflow kit, what this repo is, where each tool reads skills from, available skills and specialized agents.
glm-image-mcp-server CLAUDE.md
Claude Code instructions for ex-takashima/glm-image-mcp-server, covering claude.md - development guide, project overview, build & run commands, install dependencies and build typescript.
sound-effects-mcp CLAUDE.md
Claude Code instructions for sawa-zen/sound-effects-mcp, covering claude.md, プロジェクト概要, よく使うコマンド, 開発 and テスト.
mcp-svg-marp CLAUDE.md
Claude Code instructions for koppe-pan/mcp-svg-marp, covering mcp-svg-marp tool instructions, mcp tools available, 📊 convertsvgtomarp, 🔍 analyzesvg and automatic slide creation workflow.