MCP server for image processing and manipulation using Core Image filters and effects
che-pixel-mcp defines 23 image manipulation tools with complete input schemas and consistent descriptions. All tools have verb-noun naming (create_document, apply_filter, etc.) and parameter descriptions. However, output schemas are entirely undocumented, no tool specifies what fields are returned, forcing LLMs to guess. Descriptions are functional but brief (avg ~60 chars), lacking context on when to use each tool or what downstream operations they enable. Parameter ranges are stated in descriptions but not enforced via JSON Schema constraints (min/max). No error handling guidance, recovery hints, or idempotent operation markers. Security: no mention of permission checks, audit logging, or rate limiting. Composition is reasonable, tools are single-purpose, but no batch variants for common multi-step workflows (e.g., apply_filter_chain exists but no batch color adjustments).
Adjust brightness, contrast, saturation, hue, exposure, gamma, vibrance, highlights, shadows, temperature, tint
Apply a Core Image filter by name with parameters
Apply multiple filters sequentially
Apply Core Image auto-enhancement filters
Add bloom/glow effect
Close and release a document from memory
Apply comic book effect
No output schemas documented. Tools return results but LLMs cannot predict field names, types, or structure. Forces agents to guess what data is available for downstream operations.
No error handling guidance. Tools lack recovery hints, error classification (retryable vs fatal), or actionable error messages. LLMs cannot self-correct on failures.
Numeric parameter ranges stated in descriptions but not enforced via JSON Schema (no minimum/maximum fields). LLMs may pass out-of-range values; validation happens at runtime only.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 67 | 2026-07-28+ | v2 |
Create a new blank image document
Apply crystallize effect
Detect and highlight edges
Apply Gaussian blur
Get info about an open document
Get image metadata (dimensions, format, color space, file size)
List all open documents
Convert to monochrome with a color tint
Reduce image noise
Open an image file (PNG, JPEG, TIFF, HEIC)
Apply pixellation effect
Save document to file
Apply sepia tone
Sharpen image luminance
Apply unsharp mask for fine sharpening control
Add vignette effect
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). LLMs cannot determine which tools are safe to retry, which modify state, or which are read-only without reading descriptions.
Descriptions are brief (avg 60 chars) and lack context on when to use each tool vs alternatives. E.g., 'Apply Gaussian blur' does not explain when to use gaussian_blur vs apply_filter with CIGaussianBlur.