IBHack

You run an MCP server that can discover Python tool classes, harness Gemini AI for intelligent recommendations, and expose an HTTP endpoint to supply tool choices and code. This makes it easier to find the most relevant tools for a given task and receive complete Python code for those tools in response.

How to use

To use the IBHack MCP Server, start the server and connect a client that supports MCP endpoints. You will request tool recommendations by describing your task, and the server will return the most relevant tools along with the complete Python code for those tools. Use the client to query for recommendations like, for example, “process CSV data” or “optimize image analysis workflow,” and the server will return a concise set of tool options tailored to your description.

How to install

Prerequisites include Python and Poetry for dependency management.

# Prerequisites
# Ensure Python is installed (3.8+ is typical)
python --version

# Install Poetry
curl -sSL https://install.python-poetry.org | python3 -

# Install dependencies
poetry install

Set required environment variables for Gemini integration and optional startup directory scanning.

export GEMINI_API_KEY="your-api-key-here"
export SCAN_DIRECTORY="/path/to/your/tools"  # Optional

Run the server using one of the supported methods.

# Method 1: Direct Python execution
python server.py

# Method 2: Using Poetry
poetry run python server.py

# Method 3: Using the Poetry script
poetry run mcp-server

Configuration and runtime

By default, the server starts and serves at the HTTP endpoint http://127.0.0.1:8000/mcp. You can customize the host, port, and path by updating the startup call in your code.

mcp.run(transport="http", host="127.0.0.1", port=8000, path="/mcp")

Environment variables and runtime details

Required environment variable for Gemini integration is GEMINI_API_KEY. You may optionally set SCAN_DIRECTORY to point to a directory that contains Python tool classes for automatic discovery at startup.

export GEMINI_API_KEY="your-api-key-here"
export SCAN_DIRECTORY="/path/to/your/tools"  # Optional

Tools available through the server

The server exposes a single MCP tool for recommendations:

• recommend_tools Finds the most relevant tools for a given description using Gemini AI. It returns a list of recommended tools with complete Python code and accompanying metadata.

Available tools

recommend_tools

Returns a list of the most relevant tools for a given query description, along with complete Python code for each recommended tool and tool metadata.