> ## Documentation Index
> Fetch the complete documentation index at: https://comfyui-mcp.artokun.io/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Blog

> Notes on building comfyui-mcp — tool design, quality, and a model-highlight series on the open image & video models you can run locally.

Writing about building [comfyui-mcp](https://github.com/artokun/comfyui-mcp): tool design, schema
quality, and what we learn shipping an MCP server people actually use.

## Model highlights

A chronological tour of the open image & video models you can run locally in
ComfyUI — what each is best at, and how to set it up in one command with a
comfyui-mcp pack and the [Panel](../panel). Oldest first, ending on the current
open-weight champ, Ideogram 4.

<CardGroup cols={2}>
  <Card title="WAN 2.2" href="./wan-2.2-comfyui">
    The open-source video king — how the high/low-noise experts actually work,
    I2V/T2V, GGUF tiers, and longer videos.
  </Card>

  <Card title="Qwen-Image & Qwen-Image-Edit" href="./qwen-image-comfyui">
    20B text-to-image plus instruction editing and manual-mask inpainting, with a
    WAN 2.2 refine combo.
  </Card>

  <Card title="WAN Animate 2.2" href="./wan-animate-comfyui">
    Drive a character with a reference video — v2v character & motion replace from
    a single image.
  </Card>

  <Card title="WAN Transparent Expressions" href="./wan-transparent-comfyui">
    Alpha-channel character expression sprites (looping WebP/GIF) for SillyTavern
    and VTuber overlays.
  </Card>

  <Card title="Z-Image (Turbo & Base)" href="./z-image-comfyui">
    6B, runs under 8 GB — the low-VRAM speed king. Turbo vs Base, ControlNet, and
    your own LoRAs.
  </Card>

  <Card title="ERNIE-Image" href="./ernie-image-comfyui">
    Apache-2.0 text-to-image with razor-sharp multilingual text rendering, under
    8 GB VRAM.
  </Card>

  <Card title="ANIMA 1.0" href="./anima-comfyui">
    A \~2B anime model that generates and trains under 6 GB — Danbooru tags plus
    natural language.
  </Card>

  <Card title="LTX-2.3" href="./ltx-2.3-comfyui">
    Fast GGUF video (T2V/I2V) and the LTX Director timeline editor inside ComfyUI.
  </Card>

  <Card title="Krea 2 Turbo" href="./krea2-comfyui">
    Krea.ai's 12B open-weight text-to-image — 8-step turbo speed, Qwen3-VL
    prompting, and JSON area prompts.
  </Card>

  <Card title="Ideogram 4 — the finale" href="./ideogram-4-comfyui">
    The 9.3B open-weight king of readable in-image text and design layout, with
    JSON area prompting.
  </Card>
</CardGroup>

## Video post-production

<CardGroup cols={2}>
  <Card title="Upscaling video in ComfyUI" href="./video-upscale-comfyui">
    SeedVR2 & FlashVSR temporal restore, the downscale-first trick, the quick
    per-frame ESRGAN path, and RIFE/FILM interpolation (built into ComfyUI 0.26).
  </Card>

  <Card title="Extending video with Pusa" href="./video-extend-pusa-comfyui">
    Pusa 2.2 temporal flowmatching on the WAN 2.2 stack — continue a clip instead
    of regenerating it, chain extensions, all on your own GPU.
  </Card>
</CardGroup>

## Posts

<Card title="ComfyUI on RunPod: rent a cloud GPU" href="./runpod-comfyui">
  A first-party RunPod connector: one-tap deploy of a pod that runs YOUR exact
  custom nodes and models, an honest local⇄pod host pill, live GPU/cost status,
  and idle auto-stop — driven from the panel and the phone, no console or CLI.
  *July 21, 2026*
</Card>

<Card title="Train a FLUX LoRA in the cloud on RunPod" href="./train-lora-runpod">
  No 24 GB card? Rent one. The LoRA trainer's new SSH transport stands ai-toolkit
  up on a RunPod pod and runs the same train\_\* flow on rented hardware — dataset
  up over ssh, .safetensors back in models/loras/. *July 21, 2026*
</Card>

<Card title="Train a character LoRA by asking" href="./lora-trainer-p1">
  The panel agent drives ostris's ai-toolkit end to end through train\_\* tools —
  preflight, dataset staging, streamed progress, honest cancel — verified with a
  real FLUX.1-dev character run on a 4090. *July 20, 2026*
</Card>

<Card title="Blind mode: a privacy promise the agent can't break" href="./blind-mode-privacy">
  A user proved our "the agent never sees the image" toggle was leaky. The fix:
  enforcement at the tool-server boundary — one wrapper that withholds pixels
  from every image-returning tool, current and future, mechanically rather than
  promptably. *July 20, 2026*
</Card>

<Card title="GPT-5.6 Sol, Terra, and Luna land in the ComfyUI panel" href="./gpt-5-6-comfyui">
  The new frontier family on your ChatGPT subscription — max and ultra efforts
  with per-model ceilings from the live catalog, account-aware deprecation that
  never force-switches a runnable pin, and the SDK-pin war story behind tracking
  a fast-moving provider. *July 20, 2026*
</Card>

<Card title="Kimi K3: the biggest open-weight model ever now drives your ComfyUI canvas" href="./kimi-k3-comfyui">
  Moonshot's 2.8T-parameter K3 — third on the intelligence boards, first on
  Frontend Code Arena, weights landing July 27 — became a first-class provider
  two days after release. Why it's a big deal, the honest caveats
  (always-max thinking, \$15/M output), and the one-key setup. *July 18, 2026*
</Card>

<Card title="The 405 that wasn't: three dialects of ComfyUI-Manager" href="./manager-405-dialects">
  A debugging war story in three acts: per-operation routes vs task envelopes vs
  a legacy UI hiding under /v2 — and the frontend catchall that turns every wrong
  guess into a 405. *July 18, 2026*
</Card>

<Card title="Your conversation now survives workflow switching" href="./panel-owned-sessions">
  The conversation is the unit of continuity; the workflow is just the canvas
  target. How the orchestrator rebinds a live agent across workflow tabs — no
  respawn, so even in-memory local models keep their history. *July 17, 2026*
</Card>

<Card title="CivitAI inside ComfyUI: browse, like, and run without a download folder" href="./civitai-in-comfyui">
  A full CivitAI browser in the sidebar — lightbox with generation details,
  like/unlike with a synced likes collection, @creator search qualifiers, and
  workflows that load straight onto the live canvas. *July 17, 2026*
</Card>

<Card title="The render finished while you were making coffee: ComfyUI on your phone" href="./comfyui-mobile-app">
  The mobile companion (iOS + Android beta): same agent, same providers, CivitAI
  parity, one multi-purpose QR to pair — and tab-mirror remote control that lets
  the phone drive a desktop session. *July 16, 2026*
</Card>

<Card title="Flatten the spaghetti: making expert workflows readable" href="./flatten-workflows">
  Get/Set buses, Reroutes, and use-everywhere broadcasts resolved to direct
  links IN PLACE — layout, groups, and widgets preserved — plus the
  positional-widget bug the work exposed. *July 15, 2026*
</Card>

<Card title="The orchestrator that never goes stale" href="./self-updating-orchestrator">
  Hourly npm self-update and dev-build self-restart, coalesced to fire only when
  every agent is idle and nothing is rendering — infrastructure that updates like
  a browser, not a printer driver. *July 14, 2026*
</Card>

<Card title="We fine-tuned Gemma 4 to drive ComfyUI — and built an arena to prove it" href="./gemma4-fine-tune-arena">
  A gemma4 ladder (e2b/e4b/12b) trained on server-verified trajectories, scored
  by a best-of-3 arena of real ComfyUI tasks — including the regression we
  published against our own model, the train/serve format mismatch behind it, and
  the temperature-0 loop bug the arena couldn't see. *July 14, 2026*
</Card>

<Card title="Giving every backend eyes — and honesty when a model doesn't have them" href="./vision-per-model">
  Vision is a property of the model, not the provider. Always attempt image
  delivery, strip-and-retry once on rejection, and leave an in-history note so
  the model can never pretend it saw an attachment it didn't. *July 14, 2026*
</Card>

<Card title="Bring any LLM: free local models now drive ComfyUI" href="./local-llms-comfyui">
  Ollama, LM Studio, llama.cpp, or any OpenAI-compatible endpoint — no account,
  no API key, fully offline. The compact tool router that makes \~200 tools fit a
  4B model, the loop-breakers small models need, and the VRAM pause that lets one
  GPU run both the diffusion model and the LLM driving it. *July 14, 2026*
</Card>

<Card title="Two brains, one canvas: ComfyUI on Claude or ChatGPT" href="./comfyui-agent-claude-or-chatgpt">
  The ComfyUI sidebar agent now runs on EITHER your Claude or your ChatGPT
  subscription — pick a provider, no API keys, full canvas-driving parity. How we
  got there: an AgentBackend port, Codex over its app-server, panel tools as a
  loopback MCP, and the schema gotcha that kept eating a tool. *June 25, 2026*
</Card>

<Card title="The self-healing ComfyUI agent: it fixes the node that crashed it" href="./self-healing-comfyui-agent">
  A Wan2.2 render hard-crashes ComfyUI at 99% with a native access violation. The
  agent reads the crash dump on reload, names the culprit node, and escalates —
  update → git pull → patch → offer a PR — plus install→restart→continue and
  video-vision so it can judge its own output. *June 25, 2026*
</Card>

<Card title="The panel that talks back: an autonomous agent on your Claude subscription" href="./panel-agent-subscription">
  Three tries to get the ComfyUI panel driven by a real background agent: channel
  pushes can't wake an idle session, the `--sdk-url` transport that did everything
  got locked down, and the Claude Agent SDK's streaming input — persistent,
  injectable, interruptible, subscription-authenticated, no API key — turned out
  to be the gold standard. *June 16, 2026*
</Card>

<Card title="Installer packs that can't rot: one manifest, every install" href="./installer-packs-that-cant-rot">
  The community "1-click installer" .bat is great UX and a maintenance trap —
  Windows-only, duplicated in a separate .sh, full of model URLs that rot
  silently. How a single manifest drives both the double-click scripts and an
  MCP-native install, with CI that fails the build the moment a model link dies.
  *June 16, 2026*
</Card>

<Card title="From B to A: sharpening 70+ tool descriptions with TDQS" href="./comfyui-mcp-tdqs-case-study">
  A case study using Glama's Tool Definition Quality Score to find and fix the one weak tool that
  was capping our whole server — plus five rules any MCP author can steal. *May 25, 2026*
</Card>