Imagician

Exposes image resizing operations through the Model Context Protocol (MCP) server interface, allowing LLM agents and tools to invoke sharp's libvips-backed scaling engine with width/height/fit parameters. Implements MCP resource handlers that accept image paths or buffers and return resized outputs, enabling programmatic dimension transformation without direct library imports.

Converts images between multiple formats (JPEG, PNG, WebP, AVIF, GIF, TIFF, etc.) using sharp's codec abstraction layer, which selects optimal encoding parameters based on input/output format pairs. Exposes format conversion as MCP tools with quality/compression trade-off controls, allowing agents to choose output formats based on use-case constraints (file size, transparency support, animation).

Crops images to specified rectangular regions using coordinate-based or gravity-based (center, top-left, etc.) anchor points. Sharp's crop implementation operates on the decoded image buffer in memory, allowing sub-pixel precision and chained operations. MCP interface accepts crop parameters (x, y, width, height) or gravity keywords, enabling agents to extract regions of interest without external coordinate calculation.

Applies compression algorithms (JPEG quality reduction, PNG optimization, WebP/AVIF quality settings) to reduce file size while controlling visual degradation. Sharp exposes quality parameters (0-100 scale) that map to codec-specific compression levels. MCP tools allow agents to compress images with explicit quality targets, enabling trade-offs between file size and perceptual quality for different delivery contexts.

Extracts image metadata (dimensions, color space, DPI, EXIF tags, ICC profiles) using sharp's metadata parsing without decoding the full image. Supports EXIF orientation correction to automatically rotate images based on camera orientation tags. MCP interface exposes metadata as structured JSON, enabling agents to inspect image properties before processing or make decisions based on EXIF data.

Chains multiple image operations (resize → compress → convert format) into a single processing pipeline that executes sequentially on the decoded buffer. Sharp's fluent API allows composing operations without intermediate file writes. MCP implementation exposes batch operations as single tool calls, reducing round-trips and enabling atomic multi-step transformations that agents can invoke as a single unit.

Generates thumbnails by combining resize, crop, and format conversion operations with aspect-ratio-aware scaling. Sharp's thumbnail implementation uses a 'cover' fit mode that scales to fill a bounding box while preserving aspect ratio, then crops excess. MCP interface accepts thumbnail dimensions and returns optimized small images suitable for UI display or search result previews.

Extracts individual frames from animated formats (GIF, WebP with animation) and provides frame-level metadata (duration, disposal method). Sharp's animated image support allows iterating over frames and applying transformations per-frame. MCP tools expose frame extraction and re-encoding with frame duration control, enabling agents to manipulate animations programmatically.

Converts images between color spaces (sRGB, Adobe RGB, Lab, etc.) and applies or removes ICC color profiles. Sharp's color space support allows specifying input and output color spaces, enabling accurate color reproduction across different display contexts. MCP interface exposes color space conversion as a tool, allowing agents to ensure color consistency in multi-device workflows.

Composites watermarks, logos, or other overlay images onto source images using sharp's composite operation. Supports positioning (top-left, center, bottom-right, etc.), opacity control, and blending modes. MCP tools expose watermarking as a declarative operation, allowing agents to apply branding or copyright marks to images programmatically.