Comfyui Mcp
ComfyUI MCP: natural-language control of ComfyUI — generate images, video, and audio, and run workflows from chat.
If you already run ComfyUI locally or on a rented pod, this is the most capable community operator available — 38 tools, ControlNet and IP-Adapter conditioning, workflow DSL round-tripping, node-pack bisection, and a RunPod connector that rents and drives a GPU from the panel. The caveat is maintenance: the docs say limited maintenance has resumed and releases are scoped to verified critical fixes, with cloud users pointed to Comfy-Org's official tooling. Adopt it for local/remote control, but
Verified 5d ago · liveness 72/100 · cite: rightaichoice.com/tools/comfyui-mcp
- AI artists already running ComfyUI who want an agent to drive it end to end
- Workflow engineers automating ComfyUI pipelines through a compact round-trippable DSL
- Teams without local GPUs renting RunPod pods and training FLUX.1-dev character LoRAs from chat
- Developers who need one MCP spanning local, remote, and community cloud mode from a single config
- New ComfyUI setups — Comfy-Org ships official Comfy Cloud MCP and In-App Agent and these docs recommend them
- Complete beginners — the agent amplifies existing ComfyUI knowledge rather than teaching it
- Anyone who needs a maintained dependency with support commitments and SLAs
We scan live Reddit threads, YouTube comments, X posts, G2 reviews and other communities — and hand you an honest verdict in under a minute.
- Honest verdict, not marketing
- Real pros & cons from real users
- Attributed quotes with receipts
3 free scans · no card needed
Skip ComfyUI MCP if you're standing up agent-driven ComfyUI for the first time in 2026 — the docs point new and cloud users to Comfy-Org's official Comfy Cloud MCP and Comfy Agent, and this project is on critical-fix-only maintenance.
Renting RunPod GPU pods from the panel bills you at RunPod's rates on top of the free MCP — idle auto-stop helps, but long renders add up
The project itself is free (MIT-style community distribution via npm and GitHub), so the real cost comparison is infrastructure: this plus a local GPU costs nothing beyond hardware, a RunPod pod meters hourly on top, and Comfy Cloud mode bills the Comfy Cloud subscription — where Comfy-Org's official MCP is the supported route. Compare it against paid hosted image/video APIs you'd otherwise call per render.
In short
Comfyui Mcp — ComfyUI MCP: natural-language control of ComfyUI — generate images, video, and audio, and run workflows from chat. Best for AI artists already running ComfyUI who want an agent to drive it end to end, Workflow engineers automating ComfyUI pipelines through a compact round-trippable DSL, Teams without local GPUs renting RunPod pods and training FLUX.1-dev character LoRAs from chat. Free to use.
What's new in Comfyui Mcp
Checked 5 days agoAcross the latest 5 updates: 4 feature updates and 1 changelog entry.
July 28: a security patch, reconnect self-heal, and OpenRouter's full catalog
Path-traversal hardening on the upload tools, panel sessions that self-heal after a full reconnect instead of sticking to a dead tab, the complete OpenRouter model list, and a Windows restart_comfyui fix.
July 21: rent a GPU from the panel — RunPod connector and cloud LoRA training
A first-party RunPod connector deploys and drives a cloud GPU pod from the panel and phone with a local⇄pod host switch, live cost/GPU status, and idle auto-stop; the LoRA trainer can now train on rented pods.
July 20: GPT-5.6, the LoRA trainer lands, Blind mode gets mechanical
GPT-5.6 family support (Sol/Terra/Luna) with a max/ultra effort scale, seven train_* tools that train character LoRAs on FLUX.1-dev from chat, and a Blind toggle that mechanically gates every image-returning tool.
July 19: see what I see — viewport tools and one honest error surface
panel_view_selected and panel_view_nodes_in_viewport let the agent read what you're looking at, and panel_get_errors becomes the single error surface joining every errored node to its cause.
July 17: Kimi K3, the CivitAI browser grows up, panel-owned sessions
Moonshot's Kimi K3 joins the provider roster, the CivitAI browser gains likes and creator filters plus load-workflow-onto-canvas, and panel-owned sessions become the default so conversations survive workflow switches.
What people actually say about Comfyui Mcp — is it worth it?
We ran a structured research pass across product reviews, community discussions, and post-purchase forum threads to surface the patterns vendors won't publish themselves. Below: the recurring strengths, the hidden costs people mention most, and the cohort that consistently regrets adopting this tool.
7 mentions across 2 sources (Bluesky, GitHub) · researched Jul 5, 2026.
Average across the 2 sources that answered — each source counts once, not each post.
- +108 tools cover everything from text-to-image to model management.
- +Local-first design ensures privacy and offline capability.
- +Supports multiple backends: local, LAN, VPS, and Comfy Cloud.
- +Agent-native architecture enables complex multi-step workflows.
- +Free and open-source with active GitHub development.
- −Heavily tied to Claude Code; other agents unsupported.
- −Limited community data makes reliability assessment difficult.
- −No official support channels or documentation depth.
- −Installation hurdles for users not familiar with dev tools.
- −Destructive operations may lack adequate user prompts.
- • Potential cloud usage fees for Comfy Cloud or remote VPS
- • Time investment for setup and troubleshooting
Viability Score
How well maintained and how widely used is Comfyui Mcp? Built from what the vendor actually publishes (docs, changelog, tutorials, integrations, pricing), whether the site is live, and how much real users discuss it. How we calculate this
Last calculated: October 2026
How we score →Key Features
- Natural-language control of ComfyUI via MCP (38 documented tools)
- Text-to-image with ControlNet and IP-Adapter conditioning via generate_image
- Video and audio generation from chat
- Workflow authoring from templates: create, modify, validate, visualize
- Compact workflow DSL with round-trip conversion
- Model search and download from HuggingFace and CivitAI, plus list and remove
- Custom node install, update, snapshot, and bisect via ComfyUI-Manager
- Install/update ComfyUI, inspect the environment, start/stop the server
- Local install auto-detection on macOS, Linux, and Windows
- Remote targeting via --comfyui-url (VPS, LAN, RunPod)
- Comfy Cloud via COMFYUI_API_KEY with no local ComfyUI required
- First-party RunPod connector: deploy pods, local⇄pod host switch, live cost/GPU status, idle auto-stop
- LoRA trainer: seven train_* tools for FLUX.1-dev character LoRAs, including on rented pods
- GPT-5.6 family (Sol/Terra/Luna) with max/ultra effort scale
- Viewport tools (panel_view_selected, panel_view_nodes_in_viewport) and panel_get_errors unified error surface
About Comfyui Mcp
ComfyUI MCP is an MCP server plus Claude Code plugin that connects your AI assistant to a running ComfyUI instance — local, remote, or Comfy Cloud — so you can generate images, video and audio, author and execute workflows, manage models and custom nodes, and control the server in natural language. It's aimed at people who already run ComfyUI and want an agent to operate it, not at beginners looking for a prompt box. The vendor documents 38 tools auto-generated from their live schemas. It is not a thin prompt-to-image bridge. generate_image handles ControlNet and IP-Adapter conditioning; workflow authoring runs from templates through modify, validate, visualize, and a compact round-trippable DSL; model management searches HuggingFace and CivitAI and can download, list, and remove. Custom node packs install, update, snapshot, and bisect through ComfyUI-Manager, while install/maintain tools cover ComfyUI installation, environment inspection, and server start/stop. One config covers all three deployment shapes: auto-detected local installs and ports on macOS, Linux, and Windows, remote machines via --comfyui-url (VPS, LAN, RunPod), or Comfy Cloud through COMFYUI_API_KEY. Recent July 2026 work added a first-party RunPod connector that deploys and drives a cloud GPU pod from the panel and phone with a host switch, live cost/GPU status, and idle auto-stop, plus seven train_* LoRA trainer tools for FLUX.1-dev character LoRAs, GPT-5.6 family support (Sol/Terra/Luna) with a max/ultra effort scale, viewport tools, a single panel_get_errors error surface, and a Blind toggle gating image-returning tools. Positioning matters here. Comfy-Org now ships official agent tooling for Comfy Cloud, and the docs recommend it for cloud users who want managed setup; this project is explicitly the community MCP for local and remote ComfyUI. Maintenance is described as "limited" but resumed, with releases focused on verified critical fixes.
Behind the Verdict
The MCP server market for image generation is mostly wrappers: take a prompt, return an image. This is the other thing. It wires an assistant into a live ComfyUI instance and lets it do the unglamorous work — read what's installed, wire the graph, free VRAM, run the render, then bisect the node pack that broke. Pick it when your ComfyUI already exists and you want an agent driving it. The RunPod connector is the sharpest addition of the last quarter: deploy a pod, watch live cost and GPU status, switch hosts between local and pod, and let idle auto-stop kill the meter. Pair that with the LoRA trainer's seven train_* tools and you can produce FLUX.1-dev character LoRAs from chat without owning a GPU. Pass when you have no local install and no interest in one. Comfy-Org's official Comfy Cloud MCP and In-App Agent are the sanctioned route on Comfy Cloud, and these docs say so themselves — an unusual bit of honesty from a vendor about its own replacement. Fresh setups should start there. The maintenance signal is real and you should price it in. "Limited maintenance has resumed" with releases scoped to verified critical fixes, automatic issue reporting disabled, and a stated reassessment point when Comfy's official local panel ships publicly. Translation: this is well-built and actively patched for now, but it is not a dependency to build a product on without a migration plan. In practice it's deep enough to replace a lot of manual clicking. The workflow DSL round-trips cleanly, panel_get_errors collapses a wall of red nodes into one cause, and viewport tools let the agent see the selection you're staring at. Those are operator features, and they show up in daily use far more than any single generation capability. The honest comparison is one MCP versus two
Researching Comfyui Mcp? Get your full AI stack in 60 seconds.
Free, no signup — tell us your goal and get tools matched to your budget & existing stack.
Real-world workflow fit
Concrete scenarios for the personas Comfyui Mcp actually fits — and what changes day-one when you adopt it.
Adds the MCP server to Claude Code, lets it auto-detect the local install and port, then asks for an image with a ControlNet pose and an IP-Adapter style reference via generate_image, and asks the panel to free VRAM before the next batch.
Outcome: First image in about two minutes per the Quickstart, with no graph-editor clicking and the environment inspected by the agent rather than by hand.
Authors a workflow from a template, edits it in the compact DSL, validates and visualizes it, round-trips it back to the graph, and calls slice_workflow or strip_workflow to load a colleague's expert graph as a starting point.
Outcome: The pipeline lives in version-controllable text and still runs on the real ComfyUI instance, so hand-tuning in the GUI remains possible.
Rents a RunPod pod from the panel with live cost and GPU status, switches the host from local to pod, runs the FLUX.1-dev character LoRA training via the train_* tools, and relies on idle auto-stop between sessions.
Outcome: Heavy generation and LoRA training happen on rented compute from the chat panel, with spend visible while the pod runs.
Use Cases
- Generate images from natural language with ControlNet and IP-Adapter conditioning via generate_image
- Author, validate, and visualize complex ComfyUI workflows without opening the graph editor
- Search and download models from HuggingFace or CivitAI without leaving your terminal
- Install, update, snapshot, or bisect custom node packs through ComfyUI-Manager in conversation
- Drive a remote ComfyUI on RunPod, a VPS, or your LAN from a Claude Code sidebar panel
- Automate end-to-end image, video, and audio pipelines through scriptable MCP tools
- Train FLUX.1-dev character LoRAs from chat with the seven train_* tools
- Rent a cloud GPU pod from the panel, watch live cost, and let idle auto-stop shut it down
Models Under the Hood
as of 2026-10-18
Limitations
- The docs state maintenance is limited: releases focus on verified critical fixes, automatic issue reporting is disabled, and support will be reassessed once Comfy's official local panel ships publicly.
- You need a running ComfyUI instance (local, remote, or Comfy Cloud); it installs as an MCP server plus Claude Code plugin rather than a standalone app.
- The mobile app is beta.
- Panel agent modes, cloud GPU rental via RunPod, and LoRA training depend on external providers and their subscriptions.
- There is no SLA and no commercial support path.
as of 2026-10-02
Verification history
We have re-verified Comfyui Mcp 10 times since . Each pass re-reads the vendor's own pages and re-checks every listed field against that evidence; passes where nothing had changed are marked as such.
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-checked, vendor evidence unchanged
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
Showing the 6 most recent of 10 verification passes.
Free to cite with attribution — this page re-verifies continuously.
12-month cost
Project the real annual outlay, including the implied monthly cost when only an annual tier is published.
Vendor list price only. Add-on usage, seat overages, and contract minimums are surfaced under Hidden costs & gotchas.
Plans compared
For each published Comfyui Mcp tier: who it actually fits, and what it adds vs. the previous tier. Cross-reference the cost calculator above for projected annual outlay.
Community Edition
$0
Ideal for
Existing ComfyUI users on macOS, Linux, or Windows who want agent control of their own instance and can self-support
What this tier adds
Starting tier: free MCP server plus Claude Code plugin with all documented tools, local/remote/cloud targeting, RunPod connector, and the LoRA trainer
Where the pricing makes sense
The company stage and team size where Comfyui Mcp's pricing actually pencils out — and where peers do it cheaper.
The project itself is free (MIT-style community distribution via npm and GitHub), so the real cost comparison is infrastructure: this plus a local GPU costs nothing beyond hardware, a RunPod pod meters hourly on top, and Comfy Cloud mode bills the Comfy Cloud subscription — where Comfy-Org's official MCP is the supported route. Compare it against paid hosted image/video APIs you'd otherwise call per render.
Setup time & first value
How long it actually takes to get something useful out of Comfyui Mcp — broken out by persona, not the marketing-page minute.
Per the Quickstart, adding the server to Claude Code gets you a first image in about two minutes if ComfyUI is already installed and detected. Wiring the MCP server to an existing local or remote ComfyUI install is a short configuration pass (environment variables, transports, tokens, --comfyui-url or COMFYUI_API_KEY). Cold-starting ComfyUI itself, adding node packs, and getting a RunPod pod or
Switching to or from Comfyui Mcp
How to bring data in from common predecessors and how to get it back out — written for the switcher, not the buyer.
- →From hand-editing ComfyUI graphs: install the MCP server, let it auto-detect the local port, then recreate existing workflows through templates or slice_workflow from saved expert graphs
- →From raw ComfyUI API scripts: move your calls behind generate_image and the workflow DSL, keeping the same local, remote, or cloud endpoint
- →From other MCP servers: add comfyui-mcp alongside, point it at the same ComfyUI instance with --comfyui-url or COMFYUI_API_KEY, and route ComfyUI calls to it
- →From a RunPod web console: connect the first-party RunPod connector and manage pods, cost, and idle auto-stop from the panel instead of the browser
- ↗To Comfy-Org's official Comfy MCP: the docs recommend this for Comfy Cloud users; re-register the official server in your host and re-add your workflows
- ↗To the Comfy In-App Agent: for zero-setup cloud work, move to Comfy-Org's agent inside Comfy Cloud rather than through this MCP
- ↗To manual ComfyUI use: export workflows from the DSL back to the graph editor; nothing here locks your models or node packs to the MCP
- ↗To another MCP host: the third-party host protocol guide covers non-Claude-Code hosts if you change assistants rather than tools
Integrations
Resources & Guides
Tutorials & Learning
ComfyUI MCPチュートリアル：ClaudeやCodexをComfyUIに接続する方法 #comfyui #comfyMCP
smallzero
How to use ComfyUI MCP with Claude Code
ComfyUI
ClaudeでComfyUIが動かせるように！(ComfyUI MCP設定＋デモ)
The_Frizzy1
YouTube returned 6 videos for “Comfyui Mcp”, and we withheld 1: 1 did not mention Comfyui Mcp. Showing the 5 we can prove are about Comfyui Mcp.
Official links
Tools that pair well with Comfyui Mcp
Common stack mates teams adopt alongside Comfyui Mcp, with the specific reason each pairing earns its keep.
Kling AI
Kling AI generates native 4K AI video with synced multilingual audio and multi-keyframe shot control.
Magnific AI
Magnific is an AI creative suite that generates images, video, audio and 3D from one credit pool and a node-based canvas.
Hedra Character-3
Hedra Character-3 generates lip-synced talking-head avatar video from a photo plus audio or a script.
Featured Head-to-Head Comparisons
Comfyui Mcp vs Splice
ComfyUI MCP and Splice serve completely different creative domains: ComfyUI MCP is a free, agent-native control plane for AI image/video generation, while Splice is a paid music production sample and plugin marketplace. Your choice depends on whether you need to automate ComfyUI workflows or build beats with royalty-free sounds — they're complementary, not competing.
Comfyui Mcp vs Storyfile
If you're an AI artist or developer who already uses ComfyUI and wants agent-native control via Claude Code, ComfyUI MCP is a free, powerful choice. If you're a museum or family looking to create authentic, interactive video conversations with real people, StoryFile's cinematic capture and AI indexing is unmatched. These tools serve completely different needs—choose based on whether you need generative image pipelines or human-based conversational AI.
Comfyui Mcp vs The New Black
ComfyUI MCP is a free, agent-native control plane for ComfyUI power users who want to automate workflows via chat — ideal if you already use ComfyUI and want agent integration with Claude/ChatGPT. The New Black is a paid fashion-design specialized tool for rapid apparel concept-to-prototype. Choose ComfyUI MCP for flexible AI image generation and workflow automation; choose The New Black for purpose-built fashion design with tech pack exports.
Alternatives to Comfyui Mcp
View allKling AI
Kling AI generates native 4K AI video with synced multilingual audio and multi-keyframe shot control.
Magnific AI
Magnific is an AI creative suite that generates images, video, audio and 3D from one credit pool and a node-based canvas.
Hedra Character-3
Hedra Character-3 generates lip-synced talking-head avatar video from a photo plus audio or a script.
Frequently Asked Questions
Categories
Best-of guides
Used Comfyui Mcp? Help shape our editorial sentiment research.