MCP Tools Integration
Connect Inigami's processing capabilities directly to AI assistants like Claude through the Model Context Protocol.
What is MCP?
The Model Context Protocol (MCP) is a standard for connecting AI assistants to external tools and data sources. Inigami implements an MCP server that exposes its processing capabilities as tools that AI models can invoke directly.
This means an AI assistant like Claude can analyze images, run object detection, and perform transforms -- all by calling Inigami tools through MCP without any custom integration code.
Configuration
Add Inigami to your MCP client configuration:
1{
2 "mcpServers": {
3 "inigami": {
4 "command": "inigami-mcp-tools",
5 "env": {
6 "INIGAMI_HOST": "127.0.0.1",
7 "INIGAMI_PORT": "10210"
8 }
9 }
10 }
11}Available MCP Tools
inigami_server_status
Check that the processor is running and reachable before sending it work.
inigami_detect_objects
YOLOv4 object detection: labels, confidences and boxes, with a choice of the tiny or full model.
inigami_estimate_depth
A monocular depth map of the image, saved on the server.
inigami_segment_image
K-means segmentation in LAB colour space: a composite and one mask per region.
inigami_detect_keypoints
ORB or SIFT keypoints, strongest first, up to a limit you set.
inigami_resize_image
Resize to a width and height in pixels.
inigami_crop_image
Crop a rectangle given its corner, width and height.
inigami_rotate_image
Rotate by an angle in degrees.
inigami_filter_image
Apply an ImageMagick filter such as blur, sharpen or edge.
inigami_transform_image
Run one of the transforms by name through a single tool.
How It Works in Practice
When an AI assistant needs to process media, it calls Inigami tools through MCP:
// The assistant calls a tool
inigami_detect_objects(image_path="/srv/photos/street.jpg", model="tiny")
// inigami-mcp-tools sends the processor one line of TCP JSON
{"id": "…", "type": "image_object_detection",
"data": {"inputFile": "/srv/photos/street.jpg", "model": "tiny"}}
// and hands the assistant the processor's message and data
Image object detection completed - found 2 objects
{
"outputFile": "/tmp/mediaprocessor_20260925_174527.json",
"result": {
"model": "tiny",
"objects": [
{"class": "person", "confidence": 0.91, "bbox": [412, 180, 96, 288]},
{"class": "car", "confidence": 0.84, "bbox": [40, 260, 310, 170]}
],
"rendered_image_path": "/tmp/mediaprocessor_20260925_174527_objects.png"
}
} Zero Integration Code
MCP handles the protocol translation. No custom API clients or SDKs needed.
AI-Native
AI assistants can reason about which tools to use and chain operations intelligently.
Same Performance
MCP tools use the same C++ processing engine. No performance overhead from the protocol layer.
Connect AI to Computer Vision
Give your AI assistant the power to see and process media with Inigami MCP tools.
Setup MCP Integration