home / mcp / microsoft cognitive training mcp server

Microsoft Cognitive Training MCP Server

MCP Server generated by mcp.ag2.ai

Installation

Add the following to your MCP client configuration file.

Configuration

{
  "mcpServers": {
    "ag2-mcp-servers-custom-vision-training-client": {
      "url": "https://api.apis.guru/v2/specs/microsoft.com/cognitiveservices-Training/3.2/openapi.json",
      "headers": {
        "CONFIG": "{...}",
        "SECURITY": "YOUR_API_KEY",
        "CONFIG_PATH": "/path/to/mcp_config.json"
      }
    }
  }
}

You run an MCP server that exposes a defined OpenAPI interface so clients can interact with a machine‑readable context model. This server lets you host and consume the Cognitive Services Training API in a standardized way, enabling consistent client integrations, testing, and tooling across environments.

How to use

Create a client that communicates with the MCP server you run. You can access the HTTP endpoint for the OpenAPI interface to understand available operations and data shapes, and you can start a local stdio server to run the MCP locally for development or testing. When you start the server, you’ll be able to direct your client requests to the server’s address and have responses flow back through the MCP protocol.

How to install

Prerequisites: install Python 3.9 or newer, ensure you have the Python package manager (pip), and the uv runtime available on your system.

Step 1: Clone the MCP server repository.

Step 2: Install development dependencies.

pip install -e ".[dev]"
```

Alternatively, install using uv to run pip in editable mode:

```sh
uv pip install --editable ".[dev]"

Run the MCP server

Start the server in stdio mode to run locally. Use the following command to launch the server as a standard input/output process.

python mcp_server/main.py stdio

Configuration and environment

Configure the server using environment variables to control how it loads its configuration and security. The following variables are commonly used.

export CONFIG_PATH=/path/to/mcp_config.json
export CONFIG='{"server":"config"}'
export SECURITY='KEY=YOUR_API_KEY'

Build and test tools

The project uses a set of tooling for quality and reliability, including linting, static analysis, and tests.

ruff check
ruff format
./scripts/static-analysis.sh
./scripts/test.sh

Notes on running from source

The server’s runtime can be driven by a command line entry point that accepts different transport modes. In this workflow you start the server in stdio mode as shown above, which is suitable for local development and testing.

Available tools

lint

Run linting and formatting to ensure code quality using ruff and related scripts.

static-analysis

Perform static analysis with mypy, bandit, and semgrep to catch type issues and security concerns.

tests

Execute unit and integration tests with pytest and generate a coverage report.

build

Build and publish the MCP server using Hatch as the build system.