comfyui-mcp
Version:
Local-first, agent-native control plane for ComfyUI — MCP server + autonomous sidebar agent that drives your live graph in natural language on ANY LLM: Claude/ChatGPT/Gemini on your subscription (no API key), free local models via Ollama (fully offline),
62 lines (46 loc) • 2.86 kB
Markdown
---
description: Browse and inspect generated ComfyUI outputs
argument-hint: "Optional filter: 'last N', 'today', 'yesterday', or filename pattern"
---
# /comfy-gallery — Browse Generated Outputs
The user wants to browse images that ComfyUI has generated, inspect their metadata, and optionally re-run or modify workflows.
## Instructions
1. **Determine the output directory.** Call `get_system_stats` to check for a custom `--output-directory` argument. If not set, use the default ComfyUI output directory (typically `~/Documents/ComfyUI/output/` or the `output/` folder relative to the ComfyUI installation).
2. **List image files.** Use the Glob tool to find image files in the output directory:
- Pattern: `**/*.png` (ComfyUI saves as PNG by default)
- Also check for `**/*.jpg`, `**/*.webp` if needed
- Sort by modification time, newest first
3. **Apply filters.** The argument is: $ARGUMENTS
Parse the filter:
- **`last N`** or **`N`**: show only the N most recent images
- **`today`**: images created today (compare file dates)
- **`yesterday`**: images from yesterday
- **Filename pattern**: glob match against filenames (e.g., `ComfyUI_00123*`)
- **No filter**: default to showing the last 10 images
4. **Present image details.** For each image, show:
- Filename
- File size (human-readable, KB/MB)
- Creation/modification date and time
- Use the Read tool to display the image thumbnail if the file is a supported image format
5. **Extract workflow metadata.** When the user selects an image or asks for details:
- Use `get_workflow (action:"from_image")` to extract the embedded workflow JSON from the PNG metadata
- Show key parameters: checkpoint used, prompt, negative prompt, seed, steps, CFG, sampler, dimensions
- Offer to visualize the workflow with `visualize_workflow`
6. **Offer actions.** For any selected image, offer to:
- **Re-run**: execute the same workflow again (new seed for variation)
- **Modify and re-run**: let the user change parameters (prompt, seed, steps, etc.) and run a modified version
- **Upscale**: create an upscale workflow using the image as input
## Example
User: `/comfy-gallery last 5`
Steps:
- Find the output directory
- List all PNG files, sorted by newest
- Show the 5 most recent with filename, size, and date
- Display thumbnails
- Ask if the user wants to inspect any image's workflow
## Notes
- ComfyUI embeds workflow metadata in PNG files by default; that is what `get_workflow (action:"from_image")` reads
- JPEG and WebP files may not have embedded metadata depending on ComfyUI settings
- The output directory may contain subdirectories for different workflows or dates
- If the output directory is empty, tell the user no images have been generated yet
- Filter large galleries (100+ images) so the output stays manageable