Loading ToolBench
MCP Server for ComfyUI text to image Workflow - ComfyUI Node Integration with support for Claude Desktop, Gemini, and Claude Code
This server has severe definition quality issues. Of 2 tools, neither has adequate schema documentation visible in the provided source code. The input schemas shown are minimal (single string parameters with basic descriptions). Descriptions are extremely terse (16-18 characters for generate_image, failing the 20-character minimum). The source code excerpt is truncated and does not show the actual MCP tool registration code, only a ComfyUI node implementation. Without seeing the full fastmcp-based server implementation (promised in src/comfy_mcp_server/__init__.py but not provided), it is impossible to verify that schemas, parameter types, error handling, or any tool metadata match MCP specification. The fragment shown appears to be a ComfyUI custom node, not an MCP server definition. Per HARD SCORING RULE: if tool definitions are inferred rather than directly visible, cap at 50; if schemas are not visible, cap schema score at 0; if descriptions under 20 chars, description score capped at 0-20.
Generate an image using ComfyUI workflow
Write an image generation prompt for a provided topic
Tool descriptions critically short: 'Generate an image using ComfyUI workflow' is 18 chars, below the 20-char minimum. LLMs cannot determine context or prerequisites from such terse text.
Input schemas visible in metadata show only single required string parameters with minimal descriptions ('The prompt text for image generation'). No enums, no ranges, no format constraints. No structured output schema documented.
Source code excerpt is truncated mid-file (pyproject.toml, __main__.py, mcp_config_node.py shown, but actual MCP server implementation in src/comfy_mcp_server/__init__.py NOT provided). Cannot verify that tools are registered with fastmcp, that schemas include all required fields, or that error handling exists.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 38 | <=2025-11-25 | v2 |
| 2026-03-09 | D | 51 | - | v1 |
No error handling guidance visible. Tools that generate images or prompts are non-idempotent and may fail (network, API, resource limits). No description of what happens on failure, how to retry, or what error messages are returned.
No parameter descriptions for generate_prompt beyond 'The topic to generate an image generation prompt for'. What formats are accepted? Any length limits? Required? Optional?
Tool naming: 'generate_prompt' and 'generate_image' are generic and could be clearer. 'generate_image' does not convey that it requires ComfyUI or that the workflow is pre-configured. 'create_image_from_prompt_via_comfyui' would be more precise, though longer.
No output schema documented. generate_prompt returns a string prompt; generate_image returns what? (image path, URL, base64, file descriptor?). Without knowing output structure, LLMs cannot plan downstream tool calls or extract what they need.
generate_image marked WRITE risk but no destructiveHint annotation visible. If tools support tool annotations, destructiveHint should be applied to irreversible image generation calls so agents know to confirm before executing.