1. System Requirements

2. Building the Server

The server build process installs system dependencies, builds OpenCV from source with DNN support, and compiles the Inigami server binary. The initial setup takes 30-60 minutes due to the OpenCV build; subsequent builds are fast.

3. Running Inigami

Start the server from the top-level directory. The Makefile there provides convenience targets for the whole suite.

4. Your First Request

With the server running, send a test request using any TCP client. Here is an example using Python:

1import json
2import socket
3
4# One JSON object per line, each way. The server is loopback-only.
5with socket.create_connection(("127.0.0.1", 10210)) as sock:
6    stream = sock.makefile("rw")
7    request = {
8        "id": "test-001",
9        "type": "image_object_detection",
10        "data": {"inputFile": "/path/to/your/image.jpg", "model": "tiny"},
11    }
12    stream.write(json.dumps(request) + "\n")
13    stream.flush()
14    response = json.loads(stream.readline())
15
16print(json.dumps(response, indent=2))

You should receive a JSON response containing detected objects with class labels, confidence scores, and bounding box coordinates.

5. MCP Tools Setup

To use Inigami as MCP tools with AI assistants like Claude, add the following 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}

Once configured, AI assistants can invoke Inigami tools for object detection, image processing, and more -- all through natural language requests.