Model Context Protocol (MCP) server that extracts image metadata (EXIF, GPS, XMP, ICC, IPTC, etc.) completely offline using the exifr library.
https://github.com/stass/exif-mcpStop wrestling with external tools and network dependencies just to read image metadata. This MCP server gives your AI assistant direct access to EXIF, GPS, XMP, ICC, and IPTC data from any image format—completely offline and lightning fast.
You're probably running exiftool commands, making API calls to metadata services, or parsing binary data manually. Each approach brings its own headaches: subprocess management, network timeouts, version conflicts, or spending hours debugging byte offset calculations.
Meanwhile, you just want to ask Claude "what camera took this photo?" or "find all images shot with my 85mm lens" without jumping through hoops.
This MCP server turns image metadata into conversational data. Built on the robust exifr library, it provides 11 specialized tools that handle everything from basic EXIF reading to GPS coordinate extraction and thumbnail retrieval.
The difference: Instead of managing external processes or parsing formats yourself, you get structured metadata through simple tool calls. No subprocess overhead, no network dependencies, no version conflicts.
The PlaceSpotter reverse geolocation service uses this extensively for development and testing. When you're processing thousands of images and need reliable, fast metadata extraction, the performance difference matters.
Concrete example: Analyzing your photo library to find shooting patterns becomes a conversation:
Claude can now answer these directly by reading your image metadata instead of you writing custom scripts.
Accept images from file paths, URLs, base64 data, or raw buffers. Whether you're processing local files, analyzing web images, or working with embedded data, the same tools handle all scenarios.
Instead of parsing generic metadata dumps, use purpose-built tools:
gps-coordinates for location dataorientation for rotation informationthumbnail for embedded previewsPure JavaScript implementation means no subprocess spawning or external tool management. This translates to faster response times and more reliable operation under load.
Add this to your Claude Desktop config:
"mcpServers": {
"exif-mcp": {
"command": "node",
"args": ["/path/to/exif-mcp/dist/server.js"]
}
}
Pair it with filesystem MCP tools and you get a complete image analysis workflow. Claude can find images in directories, extract their metadata, and answer complex questions about your photo collection.
Photo Library Analysis: Understanding your shooting habits, gear usage, and location patterns through natural conversation rather than database queries.
Development Workflows: Testing image processing pipelines by quickly inspecting metadata changes without switching tools.
Batch Processing Setup: Before writing scripts to process thousands of images, prototype the logic through conversation with your AI assistant.
Debugging Image Issues: When images don't display correctly or orientation is wrong, instantly check metadata to understand what's happening.
The value isn't just reading metadata—it's making that data accessible through conversation with your AI assistant, eliminating the context switching that kills development flow.