ComfyUI health: whether it is reachable, queue size, GPU and VRAM/RAM usage, versions.

List installed models.

Args: folder: Model folder, e.g. "checkpoints", "loras", "vae", "upscale_models", "diffusion_models", "text_encoders". Omit to list every folder. search: Case-insensitive substring filter on file names.

Ask ComfyUI to unload models and free cached VRAM/RAM.

List available workflows (local files and those saved in ComfyUI), their friendly params and which preset (image/upscale/video/audio) each one backs.

Show a workflow's friendly params and every node's literal inputs, i.e. what can be changed via comfyui_run params or inputs (".").

Queue a workflow on ComfyUI.

Args: workflow: Workflow name from comfyui_workflows. params: Friendly params defined in workflows.yaml, e.g. {"prompt": "a cat", "steps": 20}. inputs: Raw node overrides, keyed ".", e.g. {"6.text": "a cat", "KSampler.cfg": 6.5}. workflow_json: An inline API-format workflow, used instead of workflow. randomize_seed: Give seed/noise_seed inputs that were not set explicitly a new random value (otherwise ComfyUI may return a cached result and skip execution). wait: Block until the job finishes (or timeout seconds) and return its outputs. timeout: Seconds to wait when wait=true (default DEFAULT_WAIT_TIMEOUT). preview: With wait=true, also return downscaled previews of the resulting images.

Recent finished jobs (newest first), including ones run from the ComfyUI web UI. Use a prompt_id from here with comfyui_save_workflow to turn that run into a workflow.

Save the exact API-format workflow of a past job as a reusable workflow, stored in ComfyUI under workflows/api/.json. This is how UI-format workflows become runnable: run it once in the ComfyUI web UI, then save it from comfyui_history.

Convert a UI-format workflow saved in ComfyUI (e.g. "comfyui:MY_FLOW.json") to API format and save it as workflows/api/.json, without having to run it first. Resolves Set/Get nodes, reroutes and bypassed nodes. Not for workflows with subgraphs.

Args: include_bypassed: Treat bypassed nodes as active (e.g. optional LoadImage slots that are bypassed in the editor but should exist in the API version).

Status of a job: pending (with queue_position), running, success, error, interrupted, or not_found. Finished jobs include their output files.

List the files produced by a finished job (images, videos, audio...). Each file has a url that supports streaming/seeking and, for media, a player page to open in a browser.

Args: prompt_id: Job id returned by comfyui_run or a preset tool. include_temp: Also include temporary preview files (PreviewImage nodes).

Return an image so it can be seen inline (downscaled to max_size px).

Either pass prompt_id (+ index among its image outputs) or an explicit filename (+ subfolder/type), e.g. to look at an uploaded input image.

Cancel a job: removes it from the queue if pending, interrupts it if running. Without prompt_id, interrupts whatever is running now.

Show the running and pending jobs, or clear all pending jobs (action="clear"; the running job is not affected).

Upload an image/audio/video into ComfyUI's input folder, for use in Load* nodes.

Args: source: An http(s) URL, a data: URI or raw base64 content.

Generate or edit images with the default image workflow (preset "image"). Without images it generates from the prompt. With images (URLs, base64, or ComfyUI input filenames) it edits them following the prompt, if the workflow supports it; refer to them in the prompt as , ... in the order given. Size: width/height in pixels, or aspect_ratio (any ratio: "16:9", "5:4", "1.5") and/or megapixels (1.0 = ~1024x1024), converted to pixels. Without any size it generates 1024x1024; when editing it uses the size of (scaled down if very large). upscale toggles the extra upscaled output when the workflow has one. Each output file has a label (e.g. "base", "upscale") when the workflow defines them.

Empty arguments keep the workflow's own values. extra sets more workflow params or raw "." overrides; workflow swaps the preset for another workflow.

Generate or edit a single asset (object, sprite, icon, character cut-out) as a PNG with a transparent background, using the default asset workflow (preset "asset"). Describe only the subject; the workflow adds the transparency instructions itself. Size is exact pixels: width/height (default 1024x1024, rounded to multiples of 16). With images and no width/height, the size of is used (scaled down if very large). Refer to the images as , ... in the prompt; transparent input images are flattened onto white first. transparent=false keeps a normal background. upscale toggles the extra upscaled output; trim (default on) also returns copies cropped to the object. Each output file has a label ("base", "upscale", "base_trimmed"...). Previews show transparency as a checkerboard.

extra sets more workflow params or raw "." overrides; workflow swaps the preset for another workflow.

Upscale an image with the default upscale workflow (preset "upscale").

image may be a URL, data URI, base64, or the name of a file already in ComfyUI's input folder (e.g. a previous output uploaded with comfyui_upload_file).

Empty arguments keep the workflow's own values. extra sets more workflow params or raw "." overrides; workflow swaps the preset for another workflow.

Generate a video with the default video workflow (preset "video"). image/image2 are start or reference images (URL, base64 or ComfyUI input filename) if the workflow uses them. Use duration (seconds) or length (frames), whichever the workflow maps. Video is slow: by default returns immediately; poll comfyui_job_status.

Empty arguments keep the workflow's own values. extra sets more workflow params or raw "." overrides; workflow swaps the preset for another workflow.

Generate audio/music with the default audio workflow (preset "audio"). prompt describes the sound/style, lyrics is for song models, duration in seconds.

Empty arguments keep the workflow's own values. extra sets more workflow params or raw "." overrides; workflow swaps the preset for another workflow.