Connect your account, then chat with AI to run tools.
Build ComfyUI image workflows with an offline node catalog, validate drafts before they hit your GPU, and queue generations on your local machine.
Try this workflow
Starter txt2img graph
Find a basic text-to-image ComfyUI template, validate the API prompt, list my checkpoints, and queue it with a scenic landscape prompt.
Opens MCPBundles Studio with this server selected. After sign-in, chat and run tools from the same thread.
Browse all toolsBuilt for
AI Artists, Creative Technologists, Local-First ML Builders
Starter txt2img graph
Covers discover → validate → list models → run in one ask.
Find a basic text-to-image ComfyUI template, validate the API prompt, list my checkpoints, and queue it with a scenic landscape prompt.
Inspect KSampler schema
Grounds agents in real object_info schemas instead of guessing.
Show me the KSampler node inputs from the knowledge index and explain how to wire model, positive, negative, and latent_image.
Fix broken wires
Offline validation catches structural mistakes before GPU time.
This ComfyUI draft fails validation because a wire references a missing node. Repair the prompt JSON and re-validate before queueing.
Do I need ComfyUI running to search nodes?
No. Searching nodes, templates, and validating drafts works from the materialized index. Queueing workflows and fetching images requires ComfyUI on your machine with Desktop connected.
Where does the node catalog come from?
MCPBundles indexes ComfyUI object_info from your local install (or a pinned snapshot) and refreshes on demand so agents query local evidence instead of live HTTP on every turn.
What workflow JSON shape does ComfyUI expect?
The HTTP API uses an API prompt map: node id strings to [class_type, inputs]. The knowledge tools document wires, output nodes, and common txt2img patterns.
Domain knowledge for ComfyUI Workflow Builder — workflow patterns, data models, and gotchas for your AI agent.
Search node schemas and starter templates offline, validate API prompt JSON, then queue workflows on the user's local ComfyUI GPU through MCPBundles Desktop.
Retry at most three times before stopping and reporting blockers.
Clear pending jobs from the local ComfyUI execution queue. Does not stop the currently running workflow — use comfyui_interrupt for that. Pair with co...
Release GPU memory on the local ComfyUI instance. Calls ComfyUI POST /free to unload models and clear execution cache on the next idle tick. Use comfy...
Read a file from the local ComfyUI install on the routed desktop machine. Paths must be relative to input/, output/, or models/. Small files return co...
Read ComfyUI execution history and outputs metadata. Pass prompt_id (returned by comfyui_run_workflow) to fetch one job's outputs, node timing, and st...
Download a generated image from the local ComfyUI server and return it for visual analysis. Pass filename, subfolder, and type exactly as reported in ...
Read ComfyUI-Manager import failure details for custom node packs. Use when a pack is installed but its node classes do not appear in comfyui_get_node...
Check whether ComfyUI-Manager is available on the local ComfyUI instance and read its download queue state. Use before comfyui_install_model — Desktop...
Poll ComfyUI-Manager for an install queued by comfyui_install_model, comfyui_install_custom_node, or comfyui_install_node_requirements. Pass save_path...
Get the input/output schema for one ComfyUI node class. Use when constructing or debugging workflow JSON for comfyui_run_workflow. Pass the internal c...
Get ComfyUI's execution queue: jobs currently running and jobs waiting to run. Use comfyui_get_status for version and hardware info; use this when you...
Check whether ComfyUI is reachable on the user's machine and summarize runtime state. Returns ComfyUI version, Python/PyTorch versions, GPU/VRAM info ...
Return a starter ComfyUI API prompt template for common workloads (SDXL txt2img, SDXL LoRA via FluxTrainer nodes, WAN T2V smoke). Use wan_t2v_native_1...
List or fetch workflow-building guides for ComfyUI API prompts and the discover-validate-run loop.
Get one node's input/output schema from the knowledge index. For custom nodes not in the index, use comfyui_get_node_info on the live instance.
Fetch one curated template including the API prompt JSON ready for comfyui_run_workflow.
Search the materialized ComfyUI node catalog (object_info). Use before building API prompt JSON for comfyui_run_workflow.
Search curated ComfyUI API prompt templates with starter graphs for common tasks.
Validate ComfyUI API prompt JSON offline: node shape, wire references, unknown classes, and missing output nodes. Run before comfyui_run_workflow.
Install a custom node pack on the local ComfyUI instance via ComfyUI-Manager. Use comfyui_search_custom_node_catalog to discover catalog entries, or p...
Detect node classes in a workflow that are not registered on the local ComfyUI instance and queue ComfyUI-Manager installs for matching custom node pa...
Download a model into the local ComfyUI install via ComfyUI-Manager (same backend as the Manager UI). All fields must match a Manager catalog entry — ...
Install or repair Python dependencies for an installed custom node pack via ComfyUI-Manager. Uses the same Manager fix queue as the UI's dependency re...
Cancel the workflow ComfyUI is currently executing. Does not remove pending queue items — use only to stop an in-progress generation.
List custom node packs installed on the local ComfyUI instance via ComfyUI-Manager. Optionally probe whether specific node classes are registered befo...
List files in the local ComfyUI input, output, temp, or models directories. Best for input datasets and models/ checkpoints. After SaveVideo or SaveIm...
List model files available to the local ComfyUI instance for a given folder (checkpoints, loras, vae, etc.). Use the returned filenames when wiring Ch...
Write a file into the local ComfyUI install on the routed desktop machine. Paths must be relative to input/, output/, or models/ — use this to upload ...
Hard-restart the local ComfyUI server through ComfyUI-Manager. Use when comfyui_free_memory cannot reclaim resident VRAM after Flux training or repeat...
Reinstall a custom node pack via ComfyUI-Manager (uninstall + install). Use when a pack is corrupted or import failures persist after comfyui_install_...
Queue a ComfyUI workflow for execution on the user's local GPU. Pass workflow as the API prompt dict (node id → [class, inputs]). Returns prompt_id an...
Search the ComfyUI-Manager custom node catalog on the local ComfyUI instance. Returns node pack metadata for comfyui_install_custom_node. Requires Com...
Search the ComfyUI-Manager model install catalog on the local ComfyUI instance. Returns entries with the exact name, type, base, save_path, filename, ...
Validate a ComfyUI workflow against the live instance: wire references, path shape, and whether each node class is registered. Use before comfyui_run_...
Wait until a queued ComfyUI workflow finishes. Default poll_mode=auto tries WebSocket progress (value/max for ETA) then falls back to HTTP poll on kee...
MCP for local ComfyUI: offline node catalog, draft validation, and GPU workflow execution through MCPBundles Desktop. It provides 34 tools that AI agents can use through the Model Context Protocol (MCP).
Add the MCPBundles server URL to your MCP client configuration (Claude Desktop, Cursor, VS Code, etc.). The URL format is: https://mcp.mcpbundles.com/bundle/comfyui-workflow-builder. Authentication is handled automatically.
ComfyUI Workflow Builder provides 34 tools that can be called by AI agents, along with a SKILL.md that gives your AI agent domain knowledge about when and how to use them.
ComfyUI Workflow Builder uses No auth required or API Key. ComfyUI requires credentials. Connect via MCPBundles and authentication is handled automatically.
No. Searching nodes, templates, and validating drafts works from the materialized index. Queueing workflows and fetching images requires ComfyUI on your machine with Desktop connected.
MCPBundles indexes ComfyUI object_info from your local install (or a pinned snapshot) and refreshes on demand so agents query local evidence instead of live HTTP on every turn.
The HTTP API uses an API prompt map: node id strings to [class_type, inputs]. The knowledge tools document wires, output nodes, and common txt2img patterns.
Connect ComfyUI Workflow Builder to any MCP client in minutes
https://mcp.mcpbundles.com/bundle/comfyui-workflow-builderThe link prefills the Add custom connector dialog — you still review the values and click Add, then Connect to complete OAuth.
ComfyUI Workflow Builder and paste the MCP URL into Remote MCP server URL.Custom connectors at claude.ai require a paid Claude plan (Pro, Max, Team, or Enterprise).
Same data, different audiences.
ComfyUI Knowledge Index — live data product, refresh cadence + sources documented.
Open datasetMore machine learning integrations you might like
Runway provides AI-powered content generation APIs for creating images, videos, and other media. Gen...
Run ComfyUI image and video generation workflows on the user's local machine. Queue API-format workf...
Originality.AI is a content detection platform that identifies AI-generated text and checks for plag...
Google Veo is Google's advanced AI video generation platform that creates high-quality videos from t...
Materialized ComfyUI node catalog from object_info plus curated API prompt templates and offline wor...
FAL AI is a machine learning platform specializing in image and video generation through AI models. ...