For AI agents — MCPBundles Directory API
For MCP client apps (human connect):
MCP connect URL: https://mcp.mcpbundles.com/bundle/comfyui
Skill page: https://www.mcpbundles.com/skills/comfyui
Tools: 41
Live probe refreshed Nov 20, 2026 · Endpoint host mcp.mcpbundles.com
What you can do with ComfyUI
Built for
AI Artists, Local Generative Media, Workflow Automation
MCPBundles publishes this directory for MCP discovery. Except where we host or operate an offering, third-party services run under their own terms. Product and company names on this page are used in a descriptive, identifying way (including under nominative fair use where applicable); they remain the property of their owners. Nothing here grants you rights in those marks, and nothing here is an offer to sell a third party's services. Terms
ComfyUI
Add to your AI client, then sign in once when you connect.
Search a ComfyUI node catalog, validate a draft, then queue it on your own GPU and pull the image back into chat.
This listing identifies a third-party integration. MCPBundles is not affiliated with, sponsored by, or endorsed by the service or its operator unless stated on this page.
Best forAI ArtistsLocal Generative MediaWorkflow Automation
Given the latest ComfyUI history entry, fetch the output image and describe what was generated.
Context to know first
Do I need ComfyUI running to look up nodes?
No. Searching the node catalog, loading starter prompts, and validating a draft work before ComfyUI is open. Queueing a workflow and fetching images still need ComfyUI running locally with MCPBundles Desktop connected.
Does this run ComfyUI in the cloud?
No. Workflows execute on your machine through ComfyUI Desktop or a local ComfyUI server. MCPBundles Desktop forwards API calls to localhost.
Which port should I use?
ComfyUI Desktop usually listens on port 8000. Command-line ComfyUI installs often use 8188. Match the port in your MCPBundles connection to the value shown in ComfyUI's server settings.
AI Skill
SKILL.md
Domain knowledge for ComfyUI — workflow patterns, data models, and gotchas for your AI agent.
ComfyUI (local)
ComfyUI is a node-graph engine for diffusion and generative media. This bundle reaches the HTTP API on the user's machine through MCPBundles Desktop — not a hosted ComfyUI cloud.
Runtime requirements
ComfyUI Desktop or a ComfyUI server must be running locally with the HTTP API enabled.
ComfyUI Desktop defaults to 127.0.0.1:8000; CLI/portable installs often use 8188.
MCPBundles Desktop must be running on the same machine so the backend can forward HTTP to localhost.
Workflow shape
ComfyUI executes API prompt JSON: a map of string node ids to [node_class, {inputs}]. Example node classes: CheckpointLoaderSimple, CLIPTextEncode, KSampler, VAEDecode, SaveImage.
Typical generation flow:
Optionally list checkpoints under the checkpoints models folder.
Validate the graph before queueing — validation reports missing_models when weights are absent; it never auto-downloads.
Manager-listed weights: search the ComfyUI-Manager catalog, then queue install with the exact catalog fields (name, type, base, save_path, filename, url).
Weights not in Manager (ACE-Step, Stable Audio, template-only HuggingFace assets): download each missing file explicitly into the correct models/{folder}/ path.
Probe one node class at a time when building graphs — the full object-info catalog is huge.
Offline catalog first. Search the materialized node catalog, load a starter API prompt, and validate the draft before ComfyUI is running. The catalog is a snapshot from the last refresh, not the user's current custom nodes or checkpoints.
Queue the graph; the response includes prompt_id.
Wait for execution history for that prompt_id until status=success and output_files is populated.
Pull each output using output_files[].fetch_tool and fetch_params — follow each entry's fetch hint (history media fetch for videos, gifs-bucket media, and audio; model file read for safetensors and other on-disk weights).
Never guess filenames from Save node filename_prefix alone — paths include ComfyUI suffixes (_00001, nested subfolders).
Upload datasets or small files under input/, output/, or models/ only (put is capped at 100 MB).
Output pull workflow (mandatory)
The root cause of “invisible” outputs is skipping history metadata and guessing paths. The correct chain:
queue workflow graph → prompt_id
wait for history → output_files[] (fetch_tool + fetch_params per entry)
fetch each output per entry metadata → bytes / spillover URL
output_files is authoritative for MCP-queued runs. Each entry includes fetch_tool, fetch_params, and path.
Local directory listing via Desktop walks folders on the user's machine — useful for browsing datasets on disk before queueing, not as a substitute for output_files after a fresh queue.
file_path is accepted as an alias for path on model file reads.
Model file reads fall back to HTTP /view when desktop disk read fails; on miss they retry exact candidate paths when disk listing finds them.
Caller-side persistence (saving bytes to a project folder) stays outside this provider. A future chained sync job may combine queue → wait → pull in one call.
LoRA training (kohya through ComfyUI)
Kohya sd-scripts runs inside ComfyUI via custom node packs (for example Kijai's ComfyUI-FluxTrainer with InitFluxLoRATraining, FluxTrainLoop, FluxTrainSave). Queue training like any other graph — there is no separate kohya server.
Training setup order:
Confirm training node classes are registered on the instance.
If missing: search the Manager custom-node catalog, install the pack, then install pip requirements for that pack.
Place checkpoint and dataset files via Manager install or file upload.
After a failed or OOM run: interrupt if needed, free GPU memory (optionally wait until free VRAM rises), reboot through Manager only when unload is not enough.
Queue the FluxTrainer workflow, poll history, pull the .safetensors output from disk.
ComfyUI's built-in TrainLoraNode is a different trainer — not kohya. Prefer FluxTrainer nodes for kohya-style training and memory options such as blocks_to_swap on constrained GPUs.
Queue, memory, and recovery
After a failed or OOM run:
Interrupt stops the in-progress job only — it does not clear pending queue items.
Clear the queue to drop pending jobs that would re-run on stale VRAM.
Free memory unloads models on the next idle tick; optional VRAM polling reads system stats from the ComfyUI server.
Manager reboot is the hard reset when the free-memory unload action is not enough.
Long jobs (LoRA training, 15–60+ minutes): prefer HTTP polling with a long timeout, or loop queue + history until the LoRA appears under models/loras/.
Validation, templates, and Manager installs
Live validation checks wires, paths, registered node classes, and missing weights on disk.
Workflow templates ship starter API prompts for common graphs (for example built-in WAN).
List local files to confirm datasets and checkpoints on disk before queueing.
Output discovery on Desktop: local directory listing on the ComfyUI output folder path walks on-disk output via MCPBundles Desktop — use before queueing to confirm datasets, not as a substitute for output_files after MCP-queued runs.
Wait-for-history can use WebSocket progress when available, then fall back to HTTP poll without failing the turn on stream parse misses.
Install missing custom nodes for unregistered classes in a workflow via Manager.
Import-fail diagnostics explain packs that installed but failed to load.
Node schema probes return registered: false when a class is not loaded.
Manager catalog and direct downloads
Model catalog search reads the Manager external-model catalog; installs use Manager's download queue (not arbitrary URLs). Idempotent when the file is already visible in the models list.
Custom-node catalog may be empty on ComfyUI Desktop until the database channel is set correctly — read Manager channel config (selected channel, db_mode) when search returns zero; probe the Manager custom-node catalog with remote mode via the generic local HTTP request escape hatch when needed.
The install snapshot lists git custom node packs with enabled/disabled flags when the installed-pack route returns sparse rows on Desktop V4 builds.
List all model folder types (not only checkpoints/loras) before searching niche weight paths such as ipadapter or diffusion_models.
Manager install rejections often return HTTP 400; use allow_error_response on that escape hatch to read the raw body. Set expect_json=false for plain-text Manager policy routes.
Git URL custom-node installs are blocked when Manager security policy is strict — lower security in Desktop settings or install from the Manager UI.
Large checkpoints may still download after install returns in_progress; poll install status or list models.
Direct HuggingFace/Civitai downloads into models/{save_dir}/ when validation reports missing_models and the weight is not catalog-listed. Use background download (default) for multi-GB files, then poll download status or list models — blocking downloads hold the MCP SSE stream open and fail after ~15–20 minutes even though the ComfyUI host may still be downloading on disk.
Gotchas
Local GPU time: Runs on the user's hardware; large graphs can take minutes.
No built-in auth: Default ComfyUI binds to localhost only. Do not expose the port to the public internet.
History outputs: Always pull via output_files from the wait-for-history step using each entry's fetch_tool. Videos/gifs/audio use history media fetch; safetensors use model file read.
node_errors: Invalid graphs return structured errors from the queue endpoint; fix inputs and retry.
Windows paths: Validation and queue normalize path separators to the host before running.
Large local files: Multi‑MB .safetensors may return a download link (spillover) instead of inline base64.
Showing 41 of 41 tools from the public preview. Connect the server to inspect the full live tool list.
Comfyui41
Comfyui API Request
write
comfyui_api_request
Call any ComfyUI or ComfyUI-Manager HTTP route on the user's local instance. Use for Manager routes not wrapped by dedicated tools (customnode/getlist with mode=remote, manager settings, security level probes). Prefer dedicated queue, workflow, and Manager install tools when they exist. Set allow_error_response=true to capture Manager 400 bodies during install debugging.
Comfyui API Request — Call any ComfyUI or ComfyUI-Manager HTTP route on the user's local instance. Use for Manager routes not wrapped by dedicated tools (customnode/getlist with mode=remote, manager settings, security level probes). Prefer dedicated queue, workflow, and Manager install tools when they exist. Set allow_error_response=true to capture Manager 400 bodies during install debugging.
Comfyui Clear Queue — Clear pending jobs from the local ComfyUI execution queue. Does not stop the currently running workflow — interrupt execution for that. Pair with GPU memory release when recovering from OOM or switching workloads.
Comfyui Download Model — Download a model file into the local ComfyUI models folder from a direct HuggingFace or Civitai URL. Use for weights not listed in ComfyUI-Manager — call explicitly after validation reports missing_models. Defaults to background=true so multi-GB downloads return in_progress immediately; poll model download status with download_id (blocking mode fails on long transfers when the Mac↔cloud MCP SSE stream times out). Returns already_installed when the models list already shows the file.
Comfyui Free Memory — Release GPU memory on the local ComfyUI instance. Calls ComfyUI POST /free to unload models and clear execution cache on the next idle tick. Interrupt execution first if a workflow is still running. Set wait_until_vram_free_mb to block until /system_stats reports enough free VRAM, or reboot through Manager when Flux or other weights stay resident.
Comfyui Get File — Read a file from the local ComfyUI install on the routed desktop machine. Paths must be relative to input/, output/, or models/. Prefer output_files fetch_tool/fetch_params from wait-for-history. Videos and gifs-bucket outputs usually need history media fetch, not disk path read. Falls back to HTTP /view when desktop disk read fails. On miss, retries score-100 candidate paths and returns ranked suggestions.
Comfyui Get History — Read ComfyUI execution history and outputs metadata. Pass prompt_id (returned when queueing a workflow) to fetch one job's status, output_files (each with fetch_tool/fetch_params and pull_hint) — same shape as a successful wait-for-history result for already-finished jobs. Omit prompt_id to list recent history (newest first, up to max_items). Fetch each output using its fetch_tool: history media fetch for videos/gifs/audio (filename/subfolder/type), disk path fetch for models and safetensors.
Frequently Asked Questions
Do I need ComfyUI running to look up nodes?
No. Searching the node catalog, loading starter prompts, and validating a draft work before ComfyUI is open. Queueing a workflow and fetching images still need ComfyUI running locally with MCPBundles Desktop connected.
Does this run ComfyUI in the cloud?
No. Workflows execute on your machine through ComfyUI Desktop or a local ComfyUI server. MCPBundles Desktop forwards API calls to localhost.
Standard connector25 free calls every 30 days, then from $19/month
$19 / month
Subject to fair usage limits
Comfyui Get Image — Download a generated image from the local ComfyUI server for inline vision in chat. Pass filename, subfolder, and type from history image outputs. PNG/JPEG/WebP only, up to 20 MB. For video, audio, or larger files use history output fetch.
Comfyui Get Import Fail Info — Read ComfyUI-Manager import failure details for custom node packs. Use when a pack is installed but its node classes do not appear in node schema probes.
Comfyui Get Install Snapshot — Read ComfyUI-Manager's current install snapshot: git custom node packs with enabled/disabled state, commit hashes, and optional pip packages. Authoritative when the installed-pack list route returns sparse rows on Desktop V4 builds.
Comfyui Get Manager Config — Read ComfyUI-Manager database channel settings: selected channel, available channel URLs, db_mode, and update policy. Use when custom-node catalog search returns zero packs or installs behave differently than expected on ComfyUI Desktop.
Comfyui Get Manager Status — Check whether ComfyUI-Manager is available on the local ComfyUI instance and read its download queue state. Use before queuing a Manager catalog model install — Desktop usually has Manager; CLI installs need --enable-manager.
Comfyui Get Model Download Status — Poll a background direct download job or confirm a model file is visible under models/{folder}. Use download_id from the in_progress download response; optionally pass models_folder and filename to check the models list.
Comfyui Get Model Install Status — Poll ComfyUI-Manager for an install queued by a model install, custom node install, or node requirements repair. Pass save_path and filename to confirm a model appears in the models list.
Comfyui Get Node Info — Get the input/output schema for one ComfyUI node class. Use when constructing or debugging workflow JSON before queueing. Pass the internal class name (not the display title), e.g. KSampler or VAEDecode.
Comfyui Get Output — Fetch a ComfyUI history output via HTTP /view using filename, subfolder, and type from output_files fetch_params. Use for videos, gifs-bucket MP4, audio, and other media referenced in history. Returns content_base64 inline or a spillover download URL. For .safetensors under models/, use disk path fetch with path instead.
Comfyui Get Queue — Get ComfyUI's execution queue: jobs currently running and jobs waiting to run. Use the status read for version and hardware info; use this when you only need queue state.
Comfyui Get Status — Check whether ComfyUI is reachable on the user's machine and summarize runtime state. Returns ComfyUI version, Python/PyTorch versions, GPU/VRAM info from /system_stats plus the current execution queue from /queue. Use this first when debugging connection issues or before queueing a workflow.
Comfyui Get Workflow Template — 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_3b for ComfyUI 0.25+ built-in WAN nodes; wan_t2v_smoke targets third-party WanVideoModelLoader packs. Parameterize the workflow, validate against the live instance, then queue for execution.
Comfyui Index Get Guides — List or fetch workflow-building guides for ComfyUI API prompts and the discover-validate-run loop.
Comfyui Index Get Node — 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.
Comfyui Index Get Template — Fetch one curated template including the API prompt JSON ready for comfyui_run_workflow.
Comfyui Index Search Nodes — Search the materialized ComfyUI node catalog (object_info). Use before building API prompt JSON for comfyui_run_workflow.
Comfyui Index Search Templates — Search curated ComfyUI API prompt templates with starter graphs for common tasks.
Comfyui Index Validate Draft — Validate ComfyUI API prompt JSON offline: node shape, wire references, unknown classes, and missing output nodes. Run before comfyui_run_workflow.
Comfyui Install Custom Node — Install a custom node pack on the local ComfyUI instance via ComfyUI-Manager. Search the custom node catalog to discover catalog entries, or pass git_url when Manager allows git installs. Kohya-backed training uses packs such as ComfyUI-FluxTrainer — queue training through a workflow after install.
Showing the first 25 tools — full list and schemas: Directory API.
ComfyUI Desktop usually listens on port 8000. Command-line ComfyUI installs often use 8188. Match the port in your MCPBundles connection to the value shown in ComfyUI's server settings.
How do I connect ComfyUI to my AI agent?
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. Authentication is handled automatically.
How many tools does ComfyUI provide?
ComfyUI provides 41 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.
What authentication does ComfyUI require?
ComfyUI uses API Key. ComfyUI requires credentials. Connect via MCPBundles and authentication is handled automatically.
MCPBundles is an independent platform built on the open Model Context Protocol standard. MCPBundles is not affiliated with, sponsored by, or endorsed by Anthropic PBC.
More in machine learning
Other MCP servers in this category from the directory index