Skip to main content
Glama
SwayamJoshi87

comfyui-mcp

by SwayamJoshi87
OverviewSchemaSourceRelated ServersDiscussionsAdmin
Python
Remote

create_workflow

Create, modify, and validate ComfyUI workflow JSON from built-in templates, then check installed node types on a running server before execution.

Instructions

Author and check ComfyUI workflow JSON. Driven by the action parameter:

  • action:"create" — Create a ready-to-run API-format workflow from a built-in template (txt2img, img2img, upscale, inpaint, controlnet, ip_adapter, ace_step_15, stable_audio_3, remove_background, ltx_video). Pure local generation — does not contact ComfyUI and has no side effects. Returns the complete workflow JSON; pass it to action:"validate" or enqueue_workflow. Unsupplied params fall back to template defaults, so the result may reference checkpoints/models that must exist on your ComfyUI server before it will execute.

  • action:"modify" — Apply modification operations to an existing workflow. Supports: set_input, add_node, remove_node, connect, insert_between. Returns the modified workflow JSON and IDs of any newly added nodes.

  • action:"validate" — Validate a workflow WITHOUT executing it. Checks for missing node types, broken connections, invalid output indices, missing models, and other issues. Returns a list of errors and warnings.

  • action:"node_info" — Query a running ComfyUI server's /object_info endpoint for installed node type definitions. Requires a reachable ComfyUI instance; results reflect that server's installed custom nodes. Use the node_type filter to inspect a specific node before composing or modifying a workflow. Default response is a STRUCTURAL summary: input/output names and type tags, with enum (dropdown) inputs collapsed to a value count — safe for context even on Loader nodes whose model dropdowns embed the entire local model list (hundreds of KB raw). Pass verbose=true (20 or fewer matches) for the complete raw definitions including every dropdown value. When more than 20 node types match, returns only a name/category list and asks you to narrow the filter.

Input Schema

TableJSON Schema
NameRequiredDescriptionDefault
actionYesWhich authoring operation to perform. "create" requires `template` (optional `params`); "modify" requires `workflow` + `operations`; "validate" requires `workflow` (optional `health`); "node_info" takes no required parameters (optional `node_type`, `verbose`, `refresh`).
healthNoaction:"validate" — Include graph-health heuristics (disconnected nodes, duplicate model loads, orphaned branches, muted/bypassed nodes, a sampler running denoise below 1.0 on an empty latent, an image-edit graph whose sampled canvas is not derived from the reference) as info/warning issues plus a structured health section. Never affects `valid`.
paramsNoaction:"create" — Template parameters; recognized keys depend on the template. txt2img: checkpoint, positive_prompt, negative_prompt, width, height, steps, cfg, seed, sampler_name, scheduler. img2img/inpaint add image_path (and mask_path for inpaint) and denoise. upscale adds upscale_model. Unknown keys are ignored; omitted keys use template defaults.
refreshNoaction:"node_info" — If true, discard the memoized /object_info snapshot and refetch live from the connected server before answering. Use after the ComfyUI server was restarted EXTERNALLY (systemd/service manager) or new model files were added out-of-band — the cache is otherwise only invalidated by MCP-managed restarts, so loader dropdowns (model lists) would remain stale for the rest of the session (#499).
verboseNoaction:"node_info" — If true, return the full raw /object_info definitions including enum dropdown values (model lists etc.) — can be hundreds of KB per Loader node, so only use it when you need the actual enum values (e.g. exact model filenames) and the filter matches few nodes. Default false: structural summary with enum value counts.
templateNoaction:"create" (REQUIRED) — Template name, one of: txt2img, img2img, upscale, inpaint, controlnet, ip_adapter, ace_step_15, stable_audio_3, remove_background, ltx_video
workflowNoComfyUI workflow JSON (as a JSON string or object). REQUIRED for action:"modify" and action:"validate". action:"validate" accepts API format or a saved UI export (nodes[]/links[]).
node_typeNoaction:"node_info" — Filter by node class_type name (case-insensitive substring match). Omit to list all available nodes.
operationsNoaction:"modify" (REQUIRED) — Array of operations to apply in order. Each has an 'op' field: set_input, add_node, remove_node, connect, or insert_between

Schema Changelog

Changes observed during successful MCP inspections.

  1. First observedv1.0.0

TDQS

A4.8/5.0
Behavior5/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

With no annotations, the description carries the full behavioral burden and does so: create is 'Pure local generation — does not contact ComfyUI and has no side effects', node_info requires a reachable server, results reflect that server's installed nodes, the memoized snapshot is only invalidated by MCP-managed restarts (#499), and verbose output 'can be hundreds of KB per Loader node'. These are non-obvious operational traits an agent needs.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness4/5

Is the description appropriately sized, front-loaded, and free of redundancy?

Long but well-structured: a one-line purpose followed by a per-action bullet for each of the four modes, with constraints front-loaded in each bullet. Nearly every sentence is load-bearing given nine parameters and four modes, though a few parentheticals (template enumeration repeated from the enum, node_info detail) could be tightened.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

No output schema exists, so the description must document returns — and it does: complete workflow JSON for create, modified JSON plus new node IDs for modify, error/warning lists for validate, and a structural summary vs. raw definitions for node_info. All prerequisites (template for create, workflow+operations for modify, server reachability for node_info) are stated.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters4/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema coverage is 100%, so the baseline is 3; the description nonetheless adds meaning beyond the schema, such as which params keys each template recognizes and the >20-match fallback that returns only a name/category list instead of the requested definitions. This is useful routing detail, though it largely parallels the schema's own descriptions.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

Opens with a specific verb+resource ('Author and check ComfyUI workflow JSON') and then decomposes it by the driving `action` parameter into four clearly-scoped modes. An agent can tell this apart from siblings like enqueue_workflow (execution) or get_workflow (retrieval) without opening the schema.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines5/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

Gives explicit when-to-use routing per action, e.g. create's output should be passed to validate or enqueue_workflow, node_info's node_type filter should be used 'before composing or modifying a workflow', and refresh should be used after an external server restart. It also states when NOT to use verbose (large filters, >20 matches).

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Deploy Server

Other Tools