ICON MCP v104

You can access ICON MCP v104 through a dedicated MCP server that exposes the ICON MCP v104 API via a standard endpoint. This server lets AI agents and tools interact with ICON MCP v104 endpoints using a consistent, protocol-friendly interface and supports quick deployment with Docker, Docker Compose, or a native Python/UV-based setup.

How to use

Connect to the MCP server using a client adapter that supports HTTP MCP endpoints. The default exposed endpoint for client connections is at http://localhost:8000/mcp/. You can discover available tools, then invoke them to perform actions against ICON MCP v104 endpoints. For example, you can determine which tools are available and then execute a specific tool with a query or command, receiving structured results from the MCP server.

Health checks help verify the server is up and responsive. Run the health check routine to confirm the MCP server is serving requests.

How to install

Prerequisites: you need a computer with Docker or Docker Compose installed, or you can run the server directly with Python and UV depending on your preferred installation method.

Option 1 — Docker (recommended) deploys a containerized MCP server for quick starts.

git clone https://github.com/Traia-IO/icon-mcp-v104-mcp-server.git
cd icon-mcp-v104-mcp-server

# Start using the provided Docker script
./run_local_docker.sh

Option 2 — Docker Compose

Prepare a local environment by setting the port, then start the services with Docker Compose.

# Create a local environment file with at least the port setting
PORT=8000

# Start the MCP server stack
docker-compose up

Option 3 — Manual installation (UV-based runtime)

Install the Python package in editable mode, then run the server using UV as the runtime.

uv pip install -e .

uv run python -m server

Available tools

example_tool

Placeholder tool (to be implemented)