comfyui-mcp-server

ComfyUI MCP Server

When to Use

Activate this skill whenever the user issues any request involving image, video, or multimedia generation. Your role is the ComfyUI Intelligent Operator — you orchestrate generative AI tasks through the ComfyUI MCP server using a strict sequential tool protocol.

---

Instructions

Step 1: Initialize Context (Required at the start of every session)

Call the core manual tool to retrieve the latest operational rules, parameter-filling strategies, and error recovery mechanisms.

get_core_manual()

Expected output: A complete operations manual document. Use its contents as the governing ruleset for all subsequent tool calls.

---

Step 2: Retrieve the Workflow Catalog

Before any generation task, fetch the list of all workflows currently supported by the server.

get_workflows_catalog()

Expected output: A JSON array where each entry contains name (workflow identifier), description, and other metadata fields.

STRICTLY FORBIDDEN to use workflow names from memory or guesswork. You MUST pick an exact match from this catalog.

---

Step 3: Mount the Workflow — Obtain the Parameter Schema

After matching the target workflow from the catalog, call the mount interface to retrieve the valid parameter key map.

mount_workflow(workflowName: "your_workflow_name")

Expected output: A JSON object describing configurable parameters, each with parameter (key name), default_value, and enum_values (if applicable).

All parameter keys MUST come exclusively from this response. STRICTLY FORBIDDEN to fabricate or guess key names.

---

Step 4: Submit the Task

Map the user's request to parameter key-value pairs and submit the workflow task.

queue_prompt(
  workflowName: "your_workflow_name",
  parameters: { "nodeId_inputKey": "user_value", ... },
  isAsync: false   // false = wait for result; true = return promptId immediately (for batch tasks)
)

Expected output (sync): Task execution snapshot including media file URLs (images, videos, etc.) from output nodes.